Je hipoteza eksponentnega časa napačna? Predavanje o nedokončani trditvi
Izvirni naslov vira: ETH ne velja
Ryan O'Donnell predstavi domnevni podeksponentni algoritem za točni #3-SAT in sam pove, da ključna konstrukcija še ni dokončana.
Video
- Le tu: 33
Kaj pravi
- Zatrjevani rezultat je podeksponentni algoritem za eksaktni #3-SAT, kar je močnejši rezultat od podeksponentnega odločitvenega algoritma in torej pomeni zatrjevano ovržbo ETH. Redčenje in prevedba, ki ohranja velikost, nalogo preneseta na permanente redkih matrik z omejeno stopnjo, s čimer naj bi izboljšali standardni algoritem s časovno zahtevnostjo v bistvu 2^n. Predlagani pristop konstruira matrični polinom A(t) nizke stopnje, katerega vrednost pri nič je vhodna matrika, pri izbranih vrednostih argumenta pa so njegove permanente učinkovito izračunljive. Hermitova interpolacija iz vrednosti in odvodov rekonstruira prvotno permanento, odvodi pa se izračunajo s permanentami podmatrik. Želene vrednosti polinoma združujejo majhne bločno diagonalne dele z napakami nizkega ranga. Njihova konstrukcija temelji na polinomu nizke stopnje, ki se na skoraj vseh Boolovih vhodih natančno ujema s parnostjo: simetrija in koncentracija omogočata interpolacijo zgolj na osrednjih plasteh Hammingove teže. Redka neujemanja nato zagotovijo nizek rang, za povezavo tega približka z nadomestnimi matrikami pa so predlagane Hadamardove transformacije. Eksponent časovne zahtevnosti naj bi bil približno n/√log n z log-log faktorji, vendar konstrukcija matrik ostaja nedokončana, O'Donnell pa priznava vrzeli v svojem razumevanju.
- O'Donnell pravi, da klasična prevedba redkega #3-SAT na permanento ohrani odgovor do znanega faktorja in ohrani velikost, zato za njegov cilj zadostuje računanje permanent redkih matrik. u13–15 · O'Donnell klasično prevedbo pripisuje Valiantu in poznejšim poenostavitvam.
- Pri matriki z vrednostmi nič in ena permanenta šteje popolna prirejanja v dvodelnem grafu, ki ga matrika predstavlja. Pri uteženih elementih vsako prirejanje prispeva produkt uteži svojih povezav. u16–18 ·
- Permanenta matrike A velikosti n×n je Σσ ∏ᵢ Aᵢ,σ(i), kjer σ preteče permutacije n indeksov. V nasprotju s formulo za determinanto ta vsota ne vsebuje faktorja za predznak permutacije. u20–21 ·
- Bernardo pojasni, da redčenje en gost primerek 3-SAT pretvori v podeksponentno število redkih primerkov, ne pa v en sam redek primerek. u40–41 · O'Donnell pravi, da že rezultat za redke primerke zadostuje za ovržbo ETH, čeprav lahko standardno redčenje poslabša količinski prihranek.
- Permanento je mogoče izračunati v času v bistvu 2^n tudi za neredke matrike velikosti n×n. O'Donnell oriše dinamično programiranje, ki shrani po en podproblem za vsako podmnožico stolpcev, namesto da bi naštevalo n! permutacij. u50–58 · V navedeni časovni zahtevnosti so polinomski faktorji izpuščeni.
- O'Donnell pravi, da je s prevedbo mogoče zagotoviti O(1) neničelnih elementov v vsaki vrstici in stolpcu matrike, katere permanento računamo, kar ustreza dvodelnemu grafu z omejeno stopnjo. u60–63 · To je močnejša zahteva od predhodne predpostavke o skupno O(n) neničelnih elementih.
- O'Donnell stopnjo D matričnega polinoma postavi na približno √log n, pri čemer izpusti log-log faktorje, za L pa vzame približno log n/log log n. Zato ima P(t) = perm(A(t)) stopnjo največ nD. u75–87 · O'Donnell v tej predstavitvi namerno izpusti nekatere log-log faktorje.
- O'Donnell prevladujoči strošek računanja odvodov oceni na 2^(n/polylog n), pomnoženo z (n/n̂)^O(n̂). Pri n̂ približno n/√log n je dobljeni eksponent približno n/√log n, pomnoženo z log-log faktorji. u149–152 · Število izračunov vrednosti odvodov v tem obračunu prispeva le zanemarljiv polinomski faktor.
- Permanenta bločno diagonalne matrike je produkt permanent njenih diagonalnih blokov. Bloki z dimenzijo največ n/polylog n zato z uporabo osnovnega algoritma za permanento omogočajo izračun v času v bistvu 2^(n/polylog n). u183–189 · Bločna struktura lahko postane vidna šele po permutaciji vrstic in stolpcev.
- O'Donnell navaja, da je permanento matrike velikosti n×n z rangom r mogoče izračunati v času n^O(r) z dinamičnim programiranjem, ki temelji na razcepu na r matrik ranga ena. u192–199 · Dinamično programiranje je le nakazano, ne pa podrobno izpeljano.
- Za vhode z vrednostmi ±1 nad realnimi števili je parnost na L bitih enaka x₁⋯xL in zahteva multilinearno stopnjo L. O'Donnell to primerja z vhodi z vrednostmi nič in ena nad poljem po modulu dva, kjer je parnost vsota stopnje ena. u228–234 ·
- Predlagani približek G zadošča Pr[G(X) ≠ parityL(X)] ≤ ε, pri čemer je ε izbran kot 1/L^100. Zato ima resničnostna tabela G − parityL največ ε2^L neničelnih vnosov. u242–247 · Na dobrih vhodih je ujemanje natančno; na izjemnih vhodih lahko G zavzame poljubno velike vrednosti, namesto da bi ostal številčno blizu parnosti.
- Srednja plast Hammingove teže L-dimenzionalne Boolove kocke vsebuje binomial(L, L/2) nizov, kar O'Donnell opiše kot približno 2^L/√L oziroma delež 1/√L kocke. u272 · Štetje velja za srednjo plast, kadar je L/2 celo število; navedeni približek izpušča konstantne faktorje.
- O'Donnell predlaga, da bi kot nadomestek za ustrezno matriko, zgrajeno iz parnosti, sestavili H diag(G) H, kjer je na diagonali realnoštevilska resničnostna tabela funkcije G, H pa je Hadamardova matrika. u294–298 ·
- O'Donnell oriše razcep matrike A, v katerem nastopi antidiagonalna matrika, nato pa bi ta faktor nadomestili s približki, izpeljanimi iz G, in tako dobili bolj strukturirane matrike. u300–302 · Konstrukcija je predstavljena le kot nedokončana skica.
- Prav to, da je permanenta polinom stopnje n, omogoča interpolacijo med različnimi matrikami kot osrednjo dokazno tehniko. u29–30 · Algebraični zapis omogoča rekonstrukcijo težko izračunljive permanente iz povezanih vrednosti.
- Hermitova interpolacija zagotovi manjkajoče informacije, kadar je točk, v katerih je izračun enostaven, premalo: vrednosti odvodov dajejo dodatne linearne omejitve za koeficiente P(t) = perm(A(t)). u113–117 · Na voljo je le približno logaritemsko število točk za izračun vrednosti, kar je precej manj od stopnje P.
- Prihranek pri času izvajanja izhaja iz tega, da je D bistveno manjši od L: interpolacija zahteva približno nD/L redov odvodov na točko oziroma približno n/√log n, če izpustimo log-log faktorje. u118–125 · Argument zahteva dovolj omejitev iz vrednosti in odvodov, da njihovo število preseže stopnjo nD.
- Hitro računanje permanent podmatrik je vključeno v načrt, ker je odvode perm(A(t)) v strukturiranih točkah mogoče izračunati s pomočjo zbirk permanent podmatrik. u171–175 · O'Donnell algebraično izpeljavo prepusti kot vajo in ni povsem prepričan, katere podmatrike natanko so vključene.
- Struktura vsote bločno diagonalne matrike in matrike nizkega ranga omogoča tudi zahtevane izračune za podmatrike, saj prehod na podmatrike ohranja bločno strukturo in ne more povečati ranga napake. u211–212 · To zagotavlja ohranjanje strukture, potrebno za del interpolacijskega načrta, ki obravnava odvode.
- O'Donnell kot izvor relativnega ranga, ki je obratna potenca log N, opredeli delež neujemanj, ki je obratna potenca L, saj je L približno log N. u250–254 · Predavanje opredeli izvor logaritemskega faktorja, ne dokonča pa pretvorbe napake približka v rang.
- Simetrija polinomsko predstavitev na Boolovi kocki prevede na enospremenljivčni problem: simetrično funkcijo je mogoče izraziti kot polinom v x₁ + ⋯ + xL, katerega stopnja je enaka stopnji večspremenljivčnega polinoma za to funkcijo. u260–266 · Ta prevedba omogoča konstrukcijo približka parnosti z enospremenljivčno interpolacijo.
- Ker so Boolovi vhodi zgoščeni blizu Hammingove teže L/2, zadostuje ujemanje s parnostjo na pasu, katerega širina raste kot √L√log L. O'Donnell se sklicuje na Chernoffovo mejo, po kateri je delež vhodov zunaj pasu L^-100. u273–275 · Razprava o širini pasu je glede konstant neformalna, poznejši ustno navedeni izrazi za stopnjo pa si nasprotujejo.
- O'Donnell domnevno ovržbo ETH predstavi kot močnejši rezultat o štetju: v podeksponentnem času je mogoče rešiti eksaktni #3-SAT, ne le odločitvene različice 3-SAT. u1–8 ·
- Načrt dokaza je konstruirati matrični polinom A(t) nizke stopnje, pri katerem je A(0) enak vhodni matriki, medtem ko imajo A(1), …, A(L) učinkovito izračunljive permanente in permanente podmatrik. Vrednosti in odvodi pri teh strukturiranih matrikah bodo omogočili rekonstrukcijo prvotne permanente. u69–99 ·
- Uporabne strukturirane matrike so vsote B + E, kjer je B bločno diagonalna matrika z majhnimi bloki, E pa ima nizek rang. O'Donnell za izračun permanente poda časovno mejo 2^O(s)n^O(r), kjer je s velikost blokov, r pa rang napake. u204–210 ·
- O'Donnell kot ključno preostalo sestavino opredeli ujemanje polinoma nizke stopnje s parnostjo: polinom stopnje približno √L, pomnožene z logaritemskimi faktorji, je lahko enak parnosti na skoraj vseh Boolovih vhodih. Znižanje stopnje bo koristilo interpolaciji, izjemni vhodi pa bodo prispevali napako nizkega ranga. u238–247 ·
- Polinom nizke stopnje se lahko natančno ujema s parnostjo na skoraj vseh Boolovih vhodih, če se z njo ujema na srednjih plasteh Hammingove teže, drugod pa sme zavzeti poljubne vrednosti. u285–288 ·
- Enospremenljivčna interpolacija zagotovi približek parnosti nizke stopnje: za predpisane vrednosti v D + 1 točkah zadostuje stopnja največ D, zato je treba interpolirati le izbrane osrednje plasti. u285–287 ·
- O'Donnell izrecno opozori, da je predavanje nepripravljen poskus razlage dokaza, katerega vseh delov ne razume. u2–3 · Rezultat, napovedan v predavanju, je treba razlikovati od v celoti izpeljanega dokaza v tem odlomku.
- O'Donnell pravi, da predlagana konstrukcija ne more zagotoviti več kot približno log n strukturiranih točk za vrednotenje. Ena sama točka, v kateri bi bilo vrednotenje zelo preprosto, bi zahtevala tako rekoč vse rede odvodov in ne bi prinesla pohitritve. u155–162 · Načrt namesto izjemno poceni vrednotenja v eni točki predvideva zmerno poceni vrednotenje v več točkah.
- Eksaktni #3-SAT naj bi dopuščal podeksponentni algoritem z interpolacijo permanent na matrikah, ki so vsota bločno diagonalne matrike in matrike nizkega ranga ter so zgrajene z uporabo ujemanja polinoma nizke stopnje s parnostjo, čeprav celotna konstrukcija ni dokončana. Kako je mogoče ujemanje polinoma nizke stopnje s parnostjo in Hermitovo interpolacijo uporabiti za eksaktno štetje rešitev #3-SAT v podeksponentnem času?
Kako utemeljuje
- thesis Claimed subexponential exact counting hinges on unfinished permanent construction How can low-degree agreement with parity and Hermite interpolation be used to compute exact #3-SAT counts in subexponential time?
- theme Reduce exact counting to sparse permanent A size-preserving reduction makes sparse permanent computation sufficient for the claimed counting result, subject to sparsification and the lecture's explicit caveats.
- theme Recover the answer through derivatives Hermite interpolation promises a subexponential algorithm when the matrix-polynomial degree is much smaller than the number of structured evaluation points.
- theme Make structured permanents efficiently computable Small diagonal blocks plus a low-rank error permit fast permanent and submatrix-permanent computation.
- theme Trade exact parity for near-total agreement Interpolating parity on central Hamming-weight layers gives a low-degree polynomial that agrees exactly on almost all Boolean inputs.
- theme Turn parity agreement into matrix structure The unfinished construction proposes Hadamard-transformed parity approximations to produce structured matrices with low-rank errors.
Vir, odlomek za odlomkom
Prepis, preveden z umetno inteligenco iz angleščine.
- u1 No, pa poskusimo. Em, prav, torej prekinjamo naš redni pouk teorije zahtevnosti, da bi poskusili dokazati, da ETH ne velja in da je SAT mogoče rešiti v podeksponentnem času. To je izšlo pred nekaj dnevi.Well, let's try it. Uh Okay, so we interrupt our regularly scheduled complexity class to try to prove that ETH is false and that uh SAT can be solved in sub-exponential time. Uh this came out a couple days ago.
0:01 · Classroom blackboard for Lecture 18: ETH False · A wide view of a dark wall-length blackboard in an otherwise empty room, with a small amount of chalk writing on the left. The writing gives a URL to a CMU "some ETH false exposition" PDF and the heading "Lecture 18: ETH False" with a claim about SAT's running time. Two grey tables in the foreground hold a blue jacket, a water bottle and a few small items. - u2 Torej to predavanje ne bo zelo dobro pripravljeno. Običajno rad vse popolno zvadim in imam pripravljene popolne zapiske ter poznam vse podrobnosti. Tu spredaj sem ves uglajen, upam, danes pa bo ravno nasprotno. Nisem veliko spal. Pravzaprav ne vem, kako bo šlo.So, this lecture will not be very well prepared. Like normally I love to like perfectly rehearse and have my perfect notes set out and I know like all the details. I'm all suave up here, I hope, but it's going to be just the opposite today. Didn't get a lot of sleep. I don't really know how this is going to go.
- u3 Ne razumem vseh delov dokaza. Ampak dobil sem dobro zamisel: to je moja trenutna, saj veste, različica pisnega dokaza, pripravljena z UI. Zato prosim vsi vzemite prenosnike. Odprite to, dajte svojemu agentu in potem, če se mi zatakne, preprosto recite: profesorju se je zataknilo pri dokazovanju tega dela, in mi pomagajte, prav?I don't understand all the parts of the proof. But I had a good idea, which is like this is my current like, you know, AI version of the written version of the proof. So, please everybody get your laptop. Like just bring this up, give it to your agent, and then like if I get stuck, just like be like the professor is stuck trying to prove this part and help me out, okay?
- u4 Prav, pa poskusimo.Okay, let's try to do it.
- u5 >> [smeh]>> [laughter]
- u6 >> Em, bomo videli, kako daleč pridemo ali kaj se zgodi. Nimam načrta. Prav, torej poskusili bomo dokazati, da je SAT v podeksponentnem času, kar je precej imenitno. To je precej boljše. Mislim, to je manjše od 1.000.01 pri poljubnem številu ničel.>> Uh we'll we'll see how far we get or what happens. I have no plan. Uh okay, so we're going to try to prove that SAT is in sub-exponential time, which is pretty awesome. Uh this is this is much better. I mean, this is smaller than 1.000.01 for any number of zeros.
- u7 Torej je res podeksponentno. Kaj še želim povedati? Aha, pravzaprav je precej boljše od tega. Oprostite, tu bi moralo pisati 3-SAT. Precej boljše je od tega. Ne bo šlo samo za to, da vam dam formulo 3-SAT v obliki 3-CNF, vi pa poveste, ali ima izpolnjujočo prireditev ali ne.So, it's really sub-exponential. Uh what else do I want to say? Oh, in fact, it's much better than this. So, um sorry, this should say 3-SAT. Uh it's much better than this. It's going to be not just uh you know, I give you a 3-SAT 3-CNF formula, you tell me if it has a satisfying assignment or not.
- u8 Natančno bomo prešteli izpolnjujoče prireditve. Dokazali bomo torej precej močnejši rezultat, da je tudi število rešitev 3-SAT, #3-SAT, mogoče izračunati v tem eks- podeksponentnem času. Prav, zato je odlično, da smo imeli prejšnjič predavanje o zahtevnosti štetja. Em, prav.We're going to exactly count the number of satisfying assignments. So, we're going to solve the much stronger result of showing that uh number of 3-SAT sharp 3-SAT can be solved in this ex- sub-exponential amount of time as well. Okay, so it's great that we had that lecture on counting uh complexity last lecture. Um Okay.
- u9 Torej to je rezultat. Aha, kot je tudi opozoril Bernardo, ki je tukaj, če ste pametni in naredite nekaj bolj prefinjenega od običajnega Cook-Levina, to pomeni tudi, da je vse, kar je rešljivo v nedeterminističnem času n, mogoče rešiti tudi v tem času.Uh So, this is the this is the result. Um Oh, as Bernardo is here also pointed out, uh if you're smart and you do something more sophisticated than the usual Cook-Levin, this also implies that uh anything solvable in non-deterministic time n can be solved in in this amount of time as well.
- u10 To je torej res odlično. Prav, dokazal vam bom to ob nekakšni blagi predpostavki. Dokazal bom ob predpostavki, da ima dana formula 3-SAT število klavzul reda n.So, that's really great. Uh Okay, so I'm going to prove it to you under what's sort of a mild assumption. I'm going to prove it to you assuming that the three-SAT formula you're given has order n clauses.
- u11 Prav, ima n spremenljivk in lahko ima, mislim, do n na tretjo klavzul, vendar bomo predpostavili, da je redka, da ima število klavzul reda n. In želim povedati, da to ni kakšna huda predpostavka. Prejšnja dela na področju finozrnate zahtevnosti bi pokazala, da lahko, če to dokažete ob predpostavki števila klavzul reda n, dokažete tudi na splošno, le da bi morda izgubili ta kvadratni koren in bi postal log n na potenco recimo 1 skozi 30 ali kaj podobnega, kar bi bilo še vedno izjemno.Okay, so it has n variables and it could have up to I guess n cubed clauses, but we're going to assume it's sparse, has order n clauses. And I want to tell you this is not like a big deal to assume this. Um Previous work in like fine-grained complexity would show that if you could prove this under the order n clauses assumption, then you could also prove it in generality, but maybe you would lose this square root and it would become log n to the power of like 1 over 30 or something, which would still be super great.
- u12 Pravzaprav ima dokaz OpenAI poleg redčenja še nekaj dodatnih trikov, ki z nečim pametnejšim pokažejo natanko ta rezultat s kvadratnim korenom n tukaj. Tega dela nisem preučil, zato ne vem, kako gre, ampak dolg je samo kakšne tri ali štiri strani.Um In fact, uh you know, the OpenAI proof uh has some extra tricks beyond the sparsification that show exactly this result with the square root n here by doing something smarter. I didn't study that part, so I don't know how it goes, but it's only like three or four pages.
- u13 Kakor koli, dokazal vam bom to različico in bomo zadovoljni. Ja, in če naredim kakšno napako, mi povejte, če jo opazite. Prav, če bi imeli naše običajno 18. predavanje, bi govorili o #P in permanentu ter o tem, da je permanent #P-poln, in videli bi tisto tri-klasično, ampak ta problem, #3-SAT, ali pravzaprav kateri koli problem štetja za, saj veste, CSP, se reducira na permanent, na katerega vas bom spomnil, na računanje permanenta.Anyway, I'm going to prove this version to you and we'll be satisfied. Um Yeah, also if I make any mistakes, just uh tell me if you know. Uh Okay, so if we were having our normal lecture 18, it would have been about uh sharp P and the permanent and uh permanent being sharp P complete, and we would have seen that three-classical, but uh this problem, sharp three-SAT, or kind of any like counting problem for like, you know, a CSP um reduces to permanent which I'll remind you about, computing permanent.
- u14 In redukcija ohranja število izpolnjujočih prireditev oziroma odgovor do fiksnega znanega faktorja ter ohranja velikost in podobno. Zato zadošča pokazati, da je tudi permanent redkih matrik n krat n s številom neničelnih elementov reda n mogoče izračunati v istem času.And the reduction is like, uh, you know, preserves the number of satisfying assignments or answer up to a fixed known factor and it it preserves the size and stuff. Um so it suffices to show that, um, permanent for sparse n by n matrices order n non-zeros is also in this same amount of time.
- u15 Prav, to je v bistvu klasična redukcija, ki jo je podal Valiant, drugi pa so jo poenostavili. Naj vas malo spomnim, kaj je permanent. Permanent je tudi, prav, prav, permanent, če imate matriko A, ste ga že videli, pravzaprav večkrat pri tem predmetu, ampak je nekakšna formula, podobna determinanti.Okay, so this is basically a classical reduction due to Valiant and made simpler by others. Um Let me remind you a little bit about what is permanent. Permanent is also, okay, okay, permanent, if you have a matrix A is seen it before, uh, several times in this class, actually, but it's like some formula that's similar to determinant.
- u16 In če matrika A vsebuje samo ničle in enice, ima zelo naravno številsko interpretacijo. Če vsebuje samo ničle in enice, si jo lahko predstavljate kot matriko bisosednosti dvodelnega grafa z n vozlišči na levi in na desni ter s povezavami glede na to, kje so tukaj ničle in enice.And if the matrix A has only zeros and ones in it, then it has a very natural counting interpretation. Uh, if it has only zeros and ones in it, you can think of it as sort of the biadjacency matrix of a bipartite graph with n vertices on the left and right and some edges according to where the zeros and ones are here.
- u17 In permanent matrike šteje popolna prirejanja v tem grafu. Prav, torej ima tudi značaj štetja nekih objektov, ki jih je a priori eksponentno mnogo. In pravzaprav, če ima matrika A še druge elemente, denimo trojko ali recimo, em, si jo lahko predstavljate kot matriko sosednosti grafa z uteženimi povezavami, kjer imajo nekatere povezave utež tri ali morda dve.And, uh, the permanent of the matrix counts the number of perfect matchings in this graph. Okay, so it also has a flavor of like, uh, you know, counting uh, a number of objects which there are a priori exponentially many. And you know, actually if the the matrix A has, uh, other entries in it like a three or like a for example, um, you can think of this like a the adjacency matrix of an edge weighted graph where some edges have weight, um, three or maybe two.
- u18 Potem je utež popolnega prirejanja produkt uteži njegovih povezav. In če so to pozitivna cela števila, si lahko to pravzaprav predstavljate tudi kot tri vzporedne povezave in tako naprej. Prav tako je v redu, če so elementi negativni. Povezave imajo lahko negativne uteži.And then the the weight of a perfect matching is the product of the the weights on the edges. And if they're positive integers, you can actually also just think of this as like three parallel edges and so forth. Um it's also fine if they're like negative entries. Uh you know, you can have negative weights on your edges.
- u19 Pravzaprav elegantna različica te redukcije morda naredi matriko z elementi, ne vem, minus ena, nič, ena in morda dve in tri ali kaj podobnega. Torej zelo majhna cela števila. Em, prav.And in fact, uh like the slick version of this reduction maybe makes uh a matrix whose entries are like I don't know, minus one, zero, one, and maybe two and three or something. So, very small integer entries. Um Okay.
- u20 Formula za permanent je tole. Vsota, in morda vidite, zakaj na splošno šteje popolna prirejanja. Vsota po vseh permutacijah sigma, ki slikajo N v N. To je tudi nekakšna posplošena diagonala v matriki.The uh the formula for permanent is um this. Sum and you can maybe see like why it's counting perfect matchings in general. Sum over all like permutations sigma mapping N to N. Uh this is also kind of like a generalized diagonal in the in the matrix.
- u21 Produkt za I od 1 do N, A I sigma I. Prav, vidite, da je enako kot determinanta, le da tukaj ni tistega neumnega faktorja minus ena s predznakom permutacije.Product I equals 1 to N A I sigma I. Okay, you see it's just like the determinant except uh there's no like stupid negative one sign of the permutation in here.
- u22 >> [prhne]>> [snorts]
- u23 >> Em, prav. Kot glavni cilj bomo pokazali ta rezultat: če računate permanent redkih matrik N krat N, lahko to storite v nekoliko podeksponentnem času. O tem bom čez trenutek povedal še nekaj, ampak ravnokar mi je prišlo na misel, da bi moral povedati tudi, ne vem.>> Um Okay. So, for the main goal, we're going to show this result that if you're computing the permanent of N by N matrices which are sparse, you can do it in slightly sub exponential time. Um I'll make a few more comments about it in uh a second, but it just occurred to me that I should also say I don't know.
- u24 Med tem predavanjem se vam bo morda zdelo malo čudno, kot da smo odleteli v vesolje. Začel sem s tem: »Pokazal vam bom, kako rešiti 3-SAT nekoliko učinkoviteje.« In vi si mislite: »Ja, imeli bomo klavzule, izpolnjujoče prireditve, resnično in neresnično in tako naprej.« Zelo kmalu pa bomo imeli matrike, permanente, range matrik.As I'm doing this like lecture, you might find like it a little bit weird like we went into outer space. Like, you know, I started by telling you, "Oh, I'm going to show you how to solve three SAT like slightly more efficiently." And you'll be like, "Yeah, there's going to be clauses, there's going to be a satisfying assignments, there's going to be like truth and false and and so forth." And like pretty soon it's going to be like we got matrices, we got permanents, we got you know, ranks of matrices.
- u25 Imeli bomo polinomsko interpolacijo, nekaj Fourierove analize ali tenzorskega zapisa. In spraševali se boste: »Kaj se dogaja? Mislil sem, da govorimo o SAT.« Ampak pravzaprav se mi zdi lepo, ker mislim, da je vsaka sestavina dokaza teoretikom zahtevnosti precej domača.Uh we got polynomial interpolation, we got like some Fourier analysis or tensor notation. And you're going to be like, "What is going on? Like, I thought we were talking about SAT here. Um but actually I think it's kind of nice because if you go through the proof, I think like every ingredient in it is quite familiar to complexity theorists.
- u26 Če ste opravili polovico podiplomskega predmeta iz zahtevnosti, kot ste jo vi, ali celo cel predmet, kot ga boste kmalu, res mislim, da boste pri vsakem delu rekli: aha, to so nekako standardne stvari, kakršne učijo pri podiplomski zahtevnosti. In lepo je, da celoten rezultat nastane preprosto tako, da te zamisli združimo na nekoliko domiseln način.Like if you've had half a course in uh graduate complexity as you have, or even a full course in graduate complexity as you will have had soon, I really think like every piece will be like, oh yeah, this is like kind of standard stuff of the stuff sort they teach you in grad complexity. And it's nice that the whole result is just formed by, you know, putting these ideas together in a slightly clever way.
- u27 Kakor koli, to je samo za takrat, ko bomo kakšnih 57 minut po začetku poskušali razumeti neko čudno polinomsko interpolacijo bločno diagonalnih matrik. Ne vem. Vnaprej sem vas opozoril, da bomo prišli do tega in da se nam ne bo ravno zdelo, da govorimo o SAT. Ja.Anyway, this is just the like, you know, like 57 minutes in when we're trying to like understand some weird block diagonal matrix polynomial interpolation. I don't know. I will have warned you in advance that we would get there and we wouldn't like really feel like we're talking about SAT. Uh yeah.
- u28 >> Torej mislim, da ima permanent to lastnost naključne samoreducibilnosti. Je to kakor koli koristno pri prehodu s 3-SAT na permanent>> So I guess the permanent is has this like self like random self-reducibility property. Is that like at all useful from a 3-SAT to a permanent
- u29 >> Ja, torej zelo bomo uporabljali dejstvo, da je permanent podan, mislim, ta lastnost samoreducibilnosti oziroma možnost izračuna permanenta ene matrike z interpolacijo permanentov drugih matrik bo v dokazu izjemno pomembna.>> Uh yeah, so we're going to heavily use that the permanent is given I mean that self-reducibility property or um ability to like compute a permanent of one matrix by interpolation of permanents of other matrices is going to be super essential in the proof.
- u30 In vse to izhaja iz dejstva, da ta formula kaže, da je permanent polinom stopnje n v elementih matrike. Zato bomo polinome precej uporabljali.And this all stems from the fact that this formula shows that the permanent is a degree n polynomial in the entries of the matrix. So we're going to use polynomials a decent amount.
- u31 >> Ne vem, ja, torej glede vaše pripombe me zanima, ali imate občutek, da je v predstavitvi SAT s CNF na primer nekaj, kar je manj primerno za dokaz, mislim, da bi ostali v nekakšnem izvornem prostoru.>> I don't Uh yeah, so the point you were making I guess I'm I'm curious if you have a feeling of whether there's something about like the SAT CNF representation for example that's like less amenable to a proof, meaning like like to to stay in that like kind of original space.
- u32 >> Ja, ne vem. Mislim, pravi teoretik algebraične zahtevnosti bi vam rekel: polinomi so tako krasni, algebra je tako krasna, vse je toliko lepše, ko,>> Yeah, I don't know. I mean like a real algebraic complexity theorist would tell you just like, oh, polynomials are so great, algebra is so great, like everything is like so much nicer when you're like,
- u33 >> [se odkašlja]>> [clears throat]
- u34 >> ne vem, preučujete zahtevnost računanja polinoma namesto neke čudne kombinatorične funkcije. Ja, ne vem. Ko sem poskušal razumeti ta dokaz, ko sem GPT pripravil do tega, da je začel, mislim, nekako sem mu rekel, poskusiva pogledati z vidika permanenta.>> I don't know, studying the complexity of computing like a polynomial rather than computing some funny combinatorial function. Um Yeah, I don't know. Uh so somehow like when uh I was trying to understand this proof, you know, once I got, you know, GPT to start, I mean, I kind of told it let's try to look at it from the permanent viewpoint.
- u35 Zdelo se mi je precej lepše. Mislim, da entropijski dokaz res poskuša ostati v svetu kombinatorike, in z njim nisem toliko napredoval. Em, prav. Ja.Like it seemed to like be much nicer to me. Um I think the entropic proof like really tries to stay in the combinatorics world and uh I didn't make as much project progress with it. Um okay. Yep.
- u36 >> Čakajte, ali je torej ta 3-SAT s številom klavzul reda n enako težak kot 3-SAT?>> Wait, so is this 3-SAT with order n clauses like the same difficulty as 3-SAT?
- u37 >> Ne čisto, ampak s klasičnimi tehnikami redčenja je znano, kako reducirati, upam, da bom prav povedal, morda Bernardo ve, kako precej učinkovito reducirati splošni 3-SAT na 3-SAT s številom klavzul reda n.>> Um not quite, but it's known how to like reduce by like classical sparsification techniques it's known I hope I get this right how to reduce, maybe Bernardo knows, like quite efficiently from general 3-SAT to 3-SAT with order n clauses.
- u38 Dovolj učinkovito, da bi lahko z malo pazljivosti iz tega rezultata za število klavzul reda n dokazali isti rezultat s slabšo potenco pri log n tukaj. Dokaz OpenAI pa opravi še nekaj prilagojenega dodatnega dela, verjetno uporabi dejanski potek dokaza, da ne izgubi niti pri tem eksponentu.So efficiently enough that like if you take a little bit of care, knowing this result for order n clauses would let you prove the same result with like a worse power on log n here. Um and in the OpenAI proof, it does some bespoke extra work, I guess using how the proof actually goes to not even lose in that exponent.
- u39 Ampak na primer zagotovo smo vedeli, vsak strokovnjak je vedel in v literaturi je znano, da bi s tem dokazom ovrgli ETH.Um but for example, like we knew for sure like every expert knew that and it's like just known in the literature that if you'd proven this, it would break ETH.
- u40 >> Morda velja povedati, da ne primerjamo enega primerka z n klavzulami in enega gostega primerka, ampak lahko en gost primerek pretvorimo v podeksponentno število redkih primerkov.>> Maybe maybe it's worth saying that the comparison is not between like one instance with like n clauses and one like dense instance, but rather that um you you can take one dense instance and convert it into a sub-exponential number of sparse instances.
- u41 >> Ja, odlično opozorilo. In to je seveda pomembno tudi pri vaši ugotovitvi o času n n, da uporabimo močnejšo obliko redukcije. Prejšnjič smo govorili o razlikah med vrstami redukcij. Močnejšo obliko redukcije, kjer uporabimo dejstvo, da ne bomo pretvorili samo v, samo ponavljam Bernarda.>> Yeah, that's a great point. For and that's of course also important in your observation here about n n time that um yeah, you use like a stronger form of reduction. We were talking about the differences of reduction types last time. A stronger form of reduction where you use the fact that you're not just going to convert to I'm just repeating Bernardo.
- u42 >> [smeh]>> [laughter]
- u43 >> En primerek redkega 3-SAT, ampak sub x, lahko si privoščimo podeksponentno mnogo primerkov. Malo se je težko navaditi, da imamo toliko časa. Naredili bomo nekaj takega: izvedli bomo ta korak. Čas izvajanja bo povečal za faktor dve na n skozi polilog n, potem pa bomo rekli: to je v redu.>> Like one uh instance of sparse the three side but like um sub x you can afford to do sub exponentially many. It's a little bit hard to like get used to like having so much time in your hands. Like we're going to do this thing where we're like well we'll do this step. It's going to blow up the running time by a factor of two to the n over poly log n but then we're going to like oh that's fine.
- u44 Povečanje časa izvajanja za dve na n skozi log na deseto n je zanemarljivo. To je čuden način razmišljanja, ampak navaditi se ga moramo. Omeniti moram še, če niste veliko razmišljali o računanju permanenta, da se vam morda zdi še bolj impresivno, da lahko permanent izračunamo v času dve na mali o od n, saj je ob pogledu na to formulo videti, da naivni algoritem za permanent potrebuje čas n fakulteta, kar je precej več od dve na n.Like blowing up the running time by two to the n over log to 10th n is like negligible. This is a weird mindset to have but we got to get into that mindset. um I should also point out that uh feeling thought much about computing permanent um You might think that this is like even more impressive that you can get it in like two to the little of n time for permanent because if you just look at this formula the naive algorithm for computing the permanent seems to take n factorial time which is much bigger than two to the n.
- u45 To je približno n na n. Zato moramo vsekakor imeti pri roki dobro znano dejstvo iz šestdesetih let, do katerega bi po običajnem vrstnem redu pri tem predmetu prišli čez, ne vem, ne prav veliko predavanj: permanent je tudi brez redkosti izračunljiv v bistvu v času DVE NA N.It's like n to the n. So one thing that we should definitely have in our pockets is like a well-known fact from the 60s that in the normal ordering of this course we would get to in I don't know not too many lectures uh which is that permanent even without the sparsity thing permanent is computable in um basically two to THE N TIME.
- u46 >> [aplavz]>> [applause]
- u47 >> ČE STE PAMETNI, LAHKO TUKAJ POSTAVITE NEKAJ TAKEGA KOT N NA ENA. To se običajno IMENUJE RYSERJEVA FORMULA.>> IF YOU'RE SMART ABOUT IT YOU CAN PUT LIKE N TO THE ONE HERE. Uh this is usually CALLED RYSER'S FORMULA.
- u48 >> [aplavz]>> [applause]
- u49 >> TO JE TISTI ČLOVEK IZ ŠESTDESETIH LET. Mislim, da je to dokazal leta 1967. Vendar lahko to naredite tudi preprosto z dinamičnim programiranjem, kar vam bom poskusil skicirati. Ryserjeva formula je eleganten način, kako to videti, lahko pa tudi malo manj elegantno.>> IT'S THE GUY FROM THE 60S. I think he proved it in 1967. um However, you can also do it just by like dynamic programming which I'll try to sketch for you. um Ryser's formula is like some slick way to see it but you can also do it in a slightly less slick way.
- u50 Prav, torej imejte to v mislih kot izhodišče: permanent lahko izračunamo v času dve na n. In poskušamo doseči nekaj malo boljšega. Za to ne potrebujemo redkosti ali česa podobnega. Naj vam na kratko pokažem, zakaj to velja.um Okay, so like you should think of it in your mind as a baseline think that like okay permanent you can do it in two to the end time. And uh so we're trying to like get it a little bit better than that. And this doesn't need sparsity or anything. Um let me just illustrate to you uh briefly why it's true.
- u51 Upam, da mi bo uspelo kar sproti. Prav, kako bi permanent izračunali na roke? Kako izračunate determinanto na, to je enako kot determinanta, samo lažje. Kako izračunate determinanto na roke? Naučijo vas različico, pri kateri vzamete ta element, rekurzivno izračunate determinanto tega in pomnožite s tem, potem pa minus ta element krat determinanta, ko prečrtate to, pa plus minus, prečrtate determinanto in tako naprej.Uh hopefully I can do this on the spot. Okay, how would you compute the permanent by hand? Well, how do you compute the determinant by It's like the same as determinant but like easier. How do you compute the determinant by hand? You know they teach you this this version where you like oh you okay you take this entry and then you like recursively compute the determinant of this and multiply by this and then like you do negative like this entry times the determinant when you cross out this and like plus minus and you cross out the determinant and so forth.
- u52 Razvoj po kofaktorjih. Permanent je natanko enak, le da ne vstavljate tistih neumnih minusov. In to je nekako vse, mislim, če minuto razmišljate o tem, je precej očitno, zakaj deluje. Pravzaprav rečete: obravnaval bom vse permutacije, pri katerih se prvi element preslika v ena, in nato rekurzivno izračunal, potem tiste, kjer se prvi element preslika v dve, in v tri in tako naprej.Cofactor expansion. Okay, permanent's exactly the same but you don't put in the like stupid minus signs. And then it's kind of all I mean if you think about it for like 1 minute it's kind of obvious why this works. I mean you're kind of just saying like oh I'm going to consider all permutations where the first element maps to one and then recursively compute and then where the first element maps to two and where it maps to three and and so forth.
- u53 Prav, to morate narediti in spet je videti, da bo trajalo n fakulteta časa. Ampak poskusili boste naslednje. Morda boste hranili veliko, v bistvu boste uporabili memoizacijo.Okay, so you have to do that and it again seems like it's going to take n factorial time. Um but uh what you're going to try to do is the following. Maybe you're going to keep a big like you're going to do memoization basically.
- u54 Najprej boste izračunali vse permanente velikosti ena krat ena v zadnji vrstici. Potem vse permanente velikosti dve krat dve v zadnjih dveh vrsticah, ne nujno iz zaporednih stolpcev. Izračunali boste, ne vem, permanent teh dveh, te matrike dve krat dve.So first you're going to compute like all the one by one permanents in the last row. And then you're going to compute all the two by two permanents in the last two rows and not necessarily contiguous. So you'll compute like I don't know the permanent of these two, this two by two matrix.
- u55 Potem vse velikosti tri krat tri v zadnjih treh vrsticah in tako naprej. Dokler niste skoraj končali. Imate vse velikosti n minus ena krat n minus ena v zadnjih n minus ena vrsticah.And then you'll do like all the three by three uh ones in the the last three rows and so forth. Um I guess until you're like almost done. You have like all the n minus one by n minus one in the last n minus one rows.
- u56 Ampak tukaj ste zbrali samo dve na, mislim, bomo videli čez trenutek. Prav. Kako naredite naslednji oziroma zadnji korak? Potrebujete to krat permanent tega, ki ste si ga že shranili, plus to krat permanent tega, ki ste si ga že shranili, in tako naprej.Uh but here like you've only collected up like two to the I mean, we'll see in a second. Okay. So, uh how do you do the next step or the final step? Well, you need to do this thing times the permanent of this, which you've already kind of memorized, and then plus this thing times the permanent of this, which you've already memorized, and so forth.
- u57 In koliko stvari shranjujete? Za vsako množico S nekako induktivno računate permanent zadnjih S stolpcev in S vrstic ter S stolpcev.And how much stuff are you like memorizing? Well, you know, I mean, for any set S, you're computing like the permanent of like the uh last S columns and S rows and S columns kind of inductively.
- u58 Torej shranjujete samo dve na S stvari, dve na n stvari, po eno za vsak S. In zato lahko to naredite v bistvu v času dve na n. To bomo potrebovali, potrebovali pa bomo tudi, in morda bom pozneje povedal več, pa bom verjetno pozneje preskočil, da lahko permanent dovolj lepe ali strukturirane matrike, čez trenutek povem, kaj to pomeni, prav tako izračunate precej hitro.So, there's only like two to the S thing two to the n things you're memorizing here, one for each S. Okay, and this is why you can do it in basically two to the n time. Okay. So, we'll need that, and we're also going to need and maybe I'll say this a bit more later, and I'll probably skip it later, but we're also going to need that uh if you have like a sufficiently nice or structured matrix, I'll say what that means in a second, you can also compute the permanent kind of fast.
- u59 Morda je permanent težko izračunati samo, če imajo elementi neko poljubno strukturo. Prav, naj vam zdaj povem celoten načrt. No, načrt moram še malo spremeniti.Maybe it's only hard to compute the permanent if like the entries have some, you know, arbitrary structure in them. Uh Okay, so let me now tell you the overall plan. Well, I have to make one more small change to the plan.
- u60 Rekel sem, da bomo permanent redkih matrik izračunali v tem času. Redke pomeni s številom neničelnih elementov reda n. Naj to redkost še malo bolj strukturiram: v vsaki vrstici in stolpcu naj bo število neničelnih elementov reda ena.I said here we're going to compute the permanent on sparse matrices in this amount of time. Sparse meaning order n non-zeros. Uh let me just make this sparsity a little bit more structured, where we have order one non-zeros in each row column.
- u61 Prav, oziroma na tej sliki, če pogledate samoOkay, or in this picture, like if you just consider uh
- u62 >> [prhne]>> [snorts]
- u63 >> kje je podpora, kje so povezave, neničelni elementi, to pomeni graf z omejeno stopnjo na obeh straneh. Morda mislite, da je to še dodatna omejitev, ampak zagotavljam vam, da jo lahko prav tako zelo preprosto vključimo v to redukcijo. Torej zadošča, da znamo narediti to.>> where you have support, where you have like edges non-zeros, it means like it's a bounded degree graph on both sides. Okay, you might think like oh, it's another further restriction, but uh I assure you like you can sneak it into this reduction very easily as well. So, um it's sufficient to be able to do this.
- u64 Prav, in v nekem smislu, torej poskusil bom zdaj napisati načrt. Celoten načrt je vzeti vhodno matriko, recimo A, ki skoraj nima strukture, in narediti bolj strukturirane različice, tako da je računanje permanenta teh različic učinkovito, torej boljše od časa 2 na n, in da nam računanje permanenta teh bolj strukturiranih različic pomaga izračunati permanent prvotne matrike A.Okay, and in some sense, uh I I so I'll try to write now what the plan The whole plan is to like take your input matrix, let's say A, which has almost no structure, and make some more structured versions of it such that um computing permanent on the more structured versions can be done efficiently, meaning like, you know, better than 2 to the n time, and computing permanent on these more structured versions helps you compute permanent on the original matrix A.
- u65 In to je že malo v tem duhu: z nekaj predhodnimi triki, še preden smo se lotili problema, smo jo naredili malo bolj strukturirano, tako da nima samo števila neničelnih elementov reda n, ampak ima v vsaki vrstici in stolpcu število neničelnih elementov reda ena. Ja.Uh and so already this is like a little bit in this spirit where we're like, well, by some pre-trickery before we even got into the problem, we made a little bit more structured so it doesn't just have like order n non-zeros, but we have order one non-zeros in each row and column. Yep.
- u66 >> Si lahko predstavljamo, da so vsi elementi ničle in enice, ali je bistveno pomembno, da so lahko negativni?>> Can they think of the entries being all zero one or it does matter substantially if they could be negative?
- u67 >> Dobro vprašanje. Pri vsem, kar bom zdaj naredil, se lahko, če vam je ljubše, pretvarjate, da so elementi ničle in enice. Ne bodo, morda lahko dosežemo, da so minus ena, nič in ena, ampak mislim, da ne moremo doseči, da so nič in ena, ne da bi redukcija nehala biti učinkovita.>> Good question. For everything I'm about to do, it's fine if you just prefer pretend the entries are zero and one. Um they won't be uh maybe they'll be like you can arrange for them to be minus one, zero, and one, but I don't think you can arrange for them to be zero and one without making this uh reduction not efficient anymore.
- u68 Ampak za predstavo je pri vsem, kar bomo počeli, čisto v redu, če se pretvarjate, da so samo ničle in enice. Prav. Torej to je načrt celotnega dokaza. In je tudi nekoliko v tistem duhu, ki sem ga prej omenil.But, psychologically, for everything we're going to do, it's fine to pretend they're just zero and one. Uh okay. So, here is the plan for the whole proof. And it kind of also has this flavor I said before.
- u69 Torej obstaja matrični polinomSo, there's uh matrix polynomial
- u70 >> [aplavz]>> [applause]
- u71 >> ali polinomska matrika. Ne vem, kako naj temu rečem. A od T. Poskusil bom z besedami povedati, kaj je to. Lahko si to predstavljate preprosto kot matriko n krat n, katere elementi so polinomi ene spremenljivke T.>> or polynomial matrix. I don't know what to call this. Uh A of T. So, I'll just try to say in words what it is. I guess one way to think of it is like you can think of this as um just an n by n matrix, but where the entries are univariate polynomials in T.
- u72 To je en način. Tukaj ima na primer 1 + T in T - T na kvadrat, kar koli. Lahko pa si jo predstavljate kot matriko A0 + T * matrika A1 + T na kvadrat * matrika A2 in tako naprej. Morda je ta pogled nekoliko boljši.That's one way to think of it. You know, it's got like 1 + T and T - T squared whatever here. Or another way to think of it is it's like a matrix A0 + T * a matrix A1 + T squared * a matrix A2 and so forth. Maybe this way is a bit better to think about it.
- u73 Tako se ne napiše polinom. A T z naslednjimi lastnostmi. Najprej, prav, tako daThat's not how you spell polynomial. Uh AT with the following properties. So, um First of all, Okay, such that um
- u74 >> [prhne]>> [snorts]
- u75 >> stopnja A T>> the degree of AT
- u76 >> [aplavz]>> [applause]
- u77 >> znaša v bistvu največ koren iz log n. Če že morate vedeti, je tukaj še log log. In veliki O, ampak kakor koli. Aha, še nekaj bi moral povedati.>> is basically at most uh root log n. Okay, I mean if you must know, there's like a log log in here. And a big O, but whatever. Um Oh, I should have said another thing.
- u78 Dokaz, ki ga bom pokazal, bo imel tukaj dva log log, medtem ko lahko dokaz OpenAI ob dovolj truda očitno doseže en log log. S tem se vsekakor ne bom ukvarjal. Prav. Stopnja je torej zelo nizka, koren iz log n. To pa zlasti pomeni, da bomo preučevali tudi permanent A T.The proof I show will get like two log logs here, whereas apparently OpenAI's proof, if you work hard, can get one log log here. I'm definitely not going to care. Um okay. So, the degree is like very low, root log n. Um In particular, though, this implies that like we're also going to be studying the permanent of AT.
- u79 In ta ima stopnjo največ n krat koren iz log n. Naj ne pišem faktorjev log log. Ker ima permanent, kot smo že omenili, stopnjo n. Če imajo zdaj vsi elementi stopnjo do korena iz log n, ima celota to stopnjo.And this has degree at most n times root log n. Okay, let me not write the log log factors. Okay, because permanent, as we mentioned before, has degree n. Now, if all the entries have degree up to root log n, the whole thing is this degree.
- u80 >> [prhne]>> [snorts]
- u81 >> A0 je A. Matrika, katere permanent poskušate izračunati. In to je morda težko izračunati, vendar bo permanente drugih A I lažje izračunati.>> Um A0 is A. The the matrix that you're trying to compute the permanent of. Um And that's maybe hard to compute, but um permanent of other AIs will be easier to compute.
- u82 Torej bo to izračunljivo v času dve na n skozi polilog n. In to za I od ena do parametra L.Um so, this is going to be computable in um two to the n over polylog n time. And this is for uh I going from one up to a parameter L.
- u83 O tem bom več povedal čez trenutek. Ampak kot sem rekel, predstavljajte si to kot izjemno hitro. To je zanemarljivo hitro v primerjavi s tem, ker bo ta polilog, ne vem, log n na deseto ali pa lahko vzamete poljubno veliko potenco logaritma. Ja?Say more about that in just a second. Um but as I said before, like you should think of this is like super fast. Like that's that's like negligibly fast compared to this cuz I don't know, this polylog will be like log n to the 10 or you can make it any large power of log. Yeah?
- u84 >> Ko rečete A I, je tam I vstavljen za T ali I vstavljen>> And you say AI, there is that I plugged in for T or I plugged
- u85 >> Ja. I vstavljen za T. Dobesedno si bomo, mislim, ja, morda je malo presenetljivo, ampak dobesedno si bomo predstavljali, da za T vstavljamo cela števila 0, 1, 2, 3, …, L. Prav. In L bo v bistvu log n.>> Yeah. I plugged in for T. We're literally going to be I mean it Yeah, it's maybe slightly surprising, but we're literally going to be imagining plugging the integer 0 1 2 3 {dot} {dot} {dot} L in for T. Uh okay. And L is going to be basically log n.
- u86 Pravzaprav bo, mislim, deljeno z log log n. Ta je nekoliko pomemben, deljeno z log log n. V bistvu log n. In tega še ne morete videti, ampak naj to poimenujem D.Uh I guess it's going to actually be divided by log log n. This one's a bit important, divided by log log n. Basically log n. And you you cannot see this yet, uh but let me call this D.
- u87 Stopnja. Ves ta prihranek, kvadratni koren iz log n, bo izviral iz dejstva, da je stopnja D za faktor koren iz log n manjša od L. Tega še ne vidite, čeprav boste kmalu nekoliko videli.Uh the degree. The whole savings here, square root log n, is going to come from the fact that uh this degree D is smaller than L by a factor of root log n. You cannot see this yet, although you'll somewhat see it soon.
- u88 Nisem še niti končal pisanja načrta, ampak v načrtu bi bilo morda to, kar želimo doseči, preprosto, če bi lahko dovolili, da je D enak L, domiselnost celotnega načrta pa bo v tem, da to dosežemo z D, ki je za koren iz log n manjši od L. Prav.But I haven't even finished writing the plan, but in the plan uh maybe what we're going to achieve here would be easy if you could allow D to be L, but the smartness of the whole plan is going to be able to achieve it with D root log n smaller than L. Um Okay.
- u89 Kot zelo drobno pripombo naj še, o, zmanjkuje mi prostora. No, to je tehnično, zato se s tem ne obremenjujte preveč, ampak permanent je izračunljiv v tem kratkem času.Uh As a very minor point let me also just Oh, I'm running out of space here. Well, uh say this is technical, so don't worry about too much, but the permanent is computable in uh this short amount of time.
- u90 TUDI ZA VSAKO PODMATRIKO A Z INDEKSOM I. MHM. S tem se ne obremenjujte preveč, bistvo je, da bo A I imel zelo lepo strukturo. Precej strukturiran bo, zato bo lažje izračunati njegov permanent.ALSO FOR ANY SUBMATRIX OF A SUB I. MHM. Don't worry about that too much, but point is like AI is going to have a very nice structure. It's going to be quite structured that makes it easier to compute its permanent.
- u91 In to lepo strukturo bodo imele tudi vse njegove podmatrike, zato bo tudi njihove permanente lahko izračunati. In zdaj zadnji del: permanent A od T je polinom v T.And like this nice structure will also be possessed by all its submatrices, so they will also be easy to have their permanent computed. And now the final bit is um Okay. Permanent of A of T is a polynomial in T.
- u92 Zanimali nas bodo njegovi odvodi po T. To je samo polinom ene spremenljivke. Preslika T v število. Zanimali nas bodo njegovi odvodi. In te odvode bomo vrednotili pri, no, pri I od 1 do L.And we're going to be interested in its derivatives with respect to T. It's just a univariate polynomial. We're going to be, you know, it maps T to a number. We're going to be interested in its derivatives. And we're going to be evaluating those derivatives at a Well, at I equals 1 through L.
- u93 In tudi te bo mogoče učinkovito izračunati. Torej K-ti odvod tega perm A T pri katerem koli T je enako I, no, katerem koli T in L, mislim.Um and these are also going to be efficiently computable. So, the Kth derivative of this perm A T at any like T equals I Well, any T and L I guess.
- u94 Je izračunljiv vIs computable in
- u95 >> [se odkašlja]>> [clears throat]
- u96 >> Izračunljiv bo v nekoliko slabšem času od tega. Torej ta čas, grozno razporejam stvari po tabli, ampak ta čas krat neki faktor, ki se slabša, ko K raste.>> Um It's going to be computable in Something a bit worse than this time. So uh This time this is terrible Board management but like this time times uh some factor that gets worse as K grows.
- u97 In ta faktor, ki se slabša, ko K raste, je približno N skozi K na K. Nekaj takega kot N nad K.And this factor that gets worse as K grows is like N over K to the K. It's something like N choose K.
- u98 >> [zavzdihne]>> [sighs]
- u99 >> Prav, in ne vem, če za K vstavite ena ali nič. Mislim, to je nekako močnejše od tega, ampak vse te odvode je vse težje izračunati. Bistvo pa je, da je tudi to število nekoliko oddaljeno od polnih dve na N, dokler je K nekoliko oddaljen od N.>> Um okay and I don't know if you plug in K to be you know one or zero. I mean You know this is sort of stronger than this but like all these derivatives they get sort of harder and harder to compute. Um But the point is like even this number is like a little bit away from The full two to the N as long as K is like a little bit away from N.
- u100 Kmalu vam bom pokazal, zakaj smo končali, ko imamo ta načrt. Kakšno vprašanje o dosedanji formulaciji načrta?And I will show you soon why Once we have this plan we're done. Um any questions about the plan statement so far?
- u101 >> Kakšen je čas v zadnjem delu?>> what's the time at the last part?
- u102 >> Ja, K-ti odvod je torej izračunljiv v času dve krat dve na N skozi polilog N. In to je, ne vem, log N na deseto, lahko pa vzamete velik polilog N. Torej toliko časa, kot potrebujete za osnovni permanent, krat faktor, ki raste s K, ko poskušate izračunati K-ti odvod, in raste precej hitro, ampak ne grozljivo hitro.>> Yeah, so the Kth derivative is computable in time Two times two to the N over poly log N. And this is like I don't know log N to the 10 or you can make it a big poly log N. So the same amount of time it takes you to compute like the base permanent times a factor that grows with the K where you're trying to compute the Kth derivative and it grows pretty fast but like not Horribly fast.
- u103 Ja.Yeah.
- u104 >> Torej zdaj delamo interpolacijo.>> So now we're doing interpolation.
- u105 >> Ja, ja, to je naslednji korak.>> Yeah, yeah that's the next step.
- u106 >> To je nekako domiseln način zapisa>> It's kind of a clever way to write like
- u107 >> Ja.>> Yeah.
- u108 >> Število mest, kjer vrednotimo, je precej manjše od faktorja>> The number of places we're evaluating is much smaller than the factor of
- u109 >> Ja. Natanko to bom zdaj povedal. Prav, morda si mislite, da ne veste, kam merim, ali pa vidite, da smo z interpolacijo skoraj končali. Prav, zdaj bomo uporabili interpolacijo.>> Yeah. That's exactly what I'm going to say now. So uh okay you might be like I don't know where he's going with this but or you might see that we're almost done uh using interpolation. Um Okay, so uh Yeah, we're going to now use interpolation.
- u110 Prav, zaradi kratkosti morda zapišimo P od T, hermitski od A od T.Um Okay, so let's let's maybe write for brevity like P of T Hermitian of A of T.
- u111 >> [aplavz]>> [applause]
- u112 >> In to je čisto običajen polinom ene spremenljivke, kakršne obravnavate v osnovni šoli. P od T. In strašno nas zanima ta vrednost. Vrednost pri nič, tukaj pa je ena, dve.>> And this is like a good old univariate polynomial such as you study in middle school. Um P of T. And uh we're dying to know this value. Value at zero and there's like one, two.
- u113 To gre do L. In stopnja tega je, kot smo rekli, če zanemarimo majhne faktorje, približno N krat koren iz log N. Pri vsej interpolaciji, ki smo jo prej delali pri predmetu, bi običajno rekli: če ima vaš polinom stopnjo D oziroma N log koren iz log N in ga lahko izračunate na toliko plus ena različnih mestih, to zadošča za določitev polinoma.This is up to L. And uh the degree of this thing they said like ignoring small factors is around N root log N. Okay. So normally for all the interpolation stuff we've been doing earlier in the class, we'd be like, "Oh, like if your polynomial is degree D or like N log root log N, you know, if you can compute it on like this many plus one different locations, then you can you know, it's sufficient to determine the polynomial.
- u114 Določite polinom in ga lahko izračunate na katerem koli drugem mestu, recimo pri nič. In samo po tem delu sodeč ni videti prav dobro, ker je mest, kjer ga lahko učinkovito izračunamo, samo L, kar je približno log N. Lahko izračunamo vse te vrednosti, ampak to je precej manj od determinante oziroma stopnje.You determine the polynomial and you can compute it at any other location like zero." And it seems like we're not in such great uh state if you just read uh this part because the number of places where we can compute it efficiently is only L, which is like log N. Um so yeah, we can compute all these values, but that's way smaller than the determinant or the degree.
- u115 Ampak izračunamo lahko tudi odvode. Na primer poznamo tudi naklon tukaj, naklon tukaj, naklon tukaj, vse te naklone. Poznamo tudi druge in tretje odvode in tako naprej. In vsak odvod nam da nekakšen podatek o polinomu.But the thing is we can also compute derivatives. So for example, we also know like the slope here, the slope here, the slope here, all these like slopes. And we also know the second derivatives and third derivatives and so forth and so on. And you know, every derivative gives us like one sort of data point about the the polynomial.
- u116 To je omejitev glede tega, kakšen je lahko polinom. In za določitev polinoma potrebujete samo strogo več omejitev kot neznank. Če malo pomislite, je vse pravzaprav linearen sistem. P je neznan v tem smislu, da ima neznane koeficiente.And this is like um a constraint on how the polynomial can be. And you just need that um there are like strictly more constraints than unknowns in order to determine the polynomial. If you think about for a moment, like everything is actually like a linear system. Like the sense in which P is unknown, it just has like unknown coefficients.
- u117 In trditev, da ima v določenih točkah določene vrednosti, je linearna omejitev zanje. Trditev, da imajo odvodi določene vrednosti, je druga linearna omejitev. Po linearni algebri torej potrebujemo več omejitev kot neznank in nekako bomo poznali P. To je nekoliko bolj prefinjena oblika interpolacije, ki jo GPT imenuje Hermitova interpolacija, ampak mislim, da je v teoretičnem računalništvu dobro znano, da lahko interpoliramo tudi z odvodi.And to say that its values at certain points are certain values is like a linear constraint on them. To say the derivatives are certain values is like another linear constraint. So, we just need, you know, by linear algebra, like more constraints than unknowns and we'll, uh, kind of know P. Uh, so this is like a slightly fancier, uh, form of interpolation, which GPT is has this called Hermite interpolation, but anyway, like I think it's like a well-known it's a well-known concept in in in TCS that you can, uh, interpolate also using derivatives.
- u118 Kaj torej potrebujemo za uspeh? Potrebujemo več podatkov, kot je neznank. Recimo, da uporabimo n s strešico odvodov.So, in order to succeed, right, what do we need? We need the amount of data we have to be bigger than the number of, uh, unknowns. So, let's say we use say we go for um, n hat many derivatives.
- u119 Torej K, oznaka odvoda tukaj, gre do n s strešico. Tako mu pravim, ker bo malo manjši od n s strešico n. Prav, skupno število omejitev bo n s strešico krat L.Okay, so K, the derivative indicator here, going up to n hat. I call n hat cuz it's going to be a little bit less than n hat n. Okay, so, um, the total number of constraints we have, uh, will be n hat times L.
- u120 Za vsako od teh L točk poznamo n s strešico odvodov ali, ne vem, n minus n s strešico minus ena, kar koli. In potrebujemo, da je to večje od stopnje, ki je n krat d, pri čemer je d ta koren iz log n.For each of these L points, we know n hat derivatives or, I don't know, n minus n hat minus one, whatever. And we need this to be bigger than the degree, uh, which is n times d, d being this, uh, root log n thing.
- u121 Ja. Torej potrebujemo, mislim, da je n s strešico skozi n vsaj d skozi L. Torej mora biti n s strešico kar precej velik delež nYeah. Okay, so I guess what that means we need that like, uh, n hat over n should be at least d over L. So, n hat has to be a pretty decently large fraction of n
- u122 >> [se odkašlja]>> [clears throat]
- u123 >> da to deluje. Vidite, zato sem rekel, da je pomembno, da je d precej manjši od L. Od tod nekako izvira prihranek. In pri nas je D opazno manjši od L. Zato je to razmerje, če zanemarimo log log, približno ena skozi koren iz log N.>> in order for this to work. Um, you see, this is where I said it's like important that d is like much smaller than L. Like somehow this is where the savings is coming from. And in our setup this D is noticeably less than L. So this ratio is ignoring log logs like around one over root log N.
- u124 To pomeni, da mora biti N s strešico večji od N skozi koren iz log N. V nekem smislu potrebujemo samo to. Torej je v redu, če je N s strešico, ne vem, do faktorjev log log samo N skozi koren iz log N.Which means that we need N hat to be bigger than N over root log N. In some sense like we only need this. So like it's okay if N hat is I don't know up to log logs merely N over root log N.
- u125 Prav, zbrati moramo samo toliko odvodov. Ja.Okay, we only need to collect this many derivatives. Yep.
- u126 >> Oprostite, zakaj je tukaj ravno N krat to?>> Sorry, why why is it N times these specifically here?
- u127 >> Aha, D je stopnja tega matričnega polinoma. Njegov permanent ima torej stopnjo N krat D. Ker ima permanent stopnjo N. Torej ko vstavite polinom stopnje T v, ja.>> Oh, uh just like D is the degree of this matrix polynomial. So the permanent of it is like N times D. Uh cuz permanent has degree N. So when you plug a degree T polynomial to Yep.
- u128 >> Torej je to preprosto dejstvo, da lahko, če imate toliko odvodov, sestavite ta sistem enačb in dobite nazaj dejansko enačbo funkcije?>> So is this just like a fact I guess that you you can if you have this many derivatives you can like set up this system of equations to then get the actual equation of the function back.
- u129 >> [se odkašlja]>> [clears throat]
- u130 >> Ja. Če koeficiente tega neznanega polinoma obravnavate kot neznanke, so vsi ti podatki o odvodih enačbe o teh koeficientih. Ko imate dovolj enačb, lahko izračunate neznanke.>> Yep. Uh if you just treat the coefficients of this unknown polynomial as unknowns then all these derivative facts are equations about these coefficients. So once you have like enough equations you can like solve for the unknowns.
- u131 Potem preprosto poznate polinom P. To zlasti pomeni, da lahko vanj vstavite nič in dobite odgovor. Ja.Then you like just simply know the polynomial P. Which in particular means you can plug zero into it and get the answer. Yep.
- u132 >> Torej N in L bistveno uporabljata dejstvo [se odkašlja], da pod- [prhne] lastnost odvodov, kajne? Dejstvo, da lahko množimo, namesto da dobimo samo plus L.>> So the N and L is critically using the fact [clears throat] that sub [snorts] derivatives property, right? Like the fact that we can multiply as opposed to just get plus L.
- u133 >> Oprostite, nisem razumel, kaj?>> Sorry, I didn't understand what?
- u134 >> Za vsak odvod, ki ga želimo izračunati, želimo dobiti še L omejitev, kajne?>> For every derivative we want to be able to compute we want to get like L more constraints, right?
- u135 >> Prav, ja.>> Okay, yeah.
- u136 >> To bi moralo uporabljati dejstvo, da lahko izračunamo L permanentov z>> That that should be using the fact that we can compute L permanents with like the
- u137 >> Ja, ja. O tem času bom govoril ravno zdaj.>> Yeah, yeah. I'll talk about this the right time just now.
- u138 >> Em>> Uh
- u139 >> No, naj govorim o>> Well, let me talk about
- u140 >> Mislim število omejitev, ki jih dobite.>> I mean like the number of constraints you're taking out.
- u141 >> Ja.>> Yeah.
- u142 >> Torej naivno, za vsak odvod imate n s strešico in potem L vrednotenj.>> So, naively like every derivative is you have n hat and then you have like L evaluations.
- u143 >> Ja, tukaj dobim n s strešico odvodov, tukaj n s strešico, tukaj n s strešico. Skupaj je to n s strešico krat L.>> Yeah, I'm getting n hat derivatives here, n hat here, n hat here. That's n hat times L all together.
- u144 >> Ja. Prav, še en [prhne]>> Yeah. Okay, another [snorts]
- u145 >> Ja.>> Yeah.
- u146 >> Em>> Um
- u147 >> O času bom govoril čez sekundo. Morda je jasno, ampak kar povejte.>> I'm going to talk about the time in just 1 second. Maybe it's clear, but go ahead.
- u148 >> Preverite. Ali to odgovori na vaše vprašanje?>> Uh Check. Does this answer your question?
- u149 >> Prav, razmislite. Dokončal bom čas. Ja. Torej vse lahko rešimo, če za vsako od L točk izračunamo n s strešico odvodov. Najprej, ni jih prav veliko. N s strešico krat L je približno n. Torej času izvajanja dodamo faktor n.>> Okay, well, think about it. I'll finish the time. Yeah. Uh Okay, so we can solve everything if we can compute n hat many derivatives for each uh of L points. Now, first of all, there's I mean, there's not that many. N hat times L is like around n. So, it's like adding a factor n to our running time.
- u150 To je povsem zanemarljivo. Glavna težava je izračun katerega koli od teh odvodov. Vse dražji so, večji kot je K, ampak tukaj lahko preprosto vstavimo K enako n s strešico, da dobimo glavni čas izvajanja. Torej sem skoraj končal.That's super negligible. Um so, the main difficulty is computing any one of these derivatives. And, you know, they get more and more expensive the bigger K is here, but we can just plug K equals n hat in here to get like the main uh running time. So, I'm almost done here.
- u151 Čas izvajanja je odvisen predvsem od primera, ko je K enak n s strešico, torej je dve na n skozi polilog n. To je zanemarljivo. Krat n skozi n s strešico na potenco reda n s strešico.Uh The running time is dominated by when K is n hat, so it's it's 2 to the n over poly log n. That's negligible. Uh times this like n over n hat to the power of order n hat.
- u152 In zdaj nam gre odlično. N skozi n s strešico je približno logaritemsko. Ko to prenesete v eksponent, je dvojno logaritemsko. Torej je to v bistvu dve na red n s strešico krat nekaj faktorjev log log. In n s strešico je n skozi koren iz log n.And, now we're in great shape. N over n hat is like logarithmic. So, like when you move to the exponent, it's like doubly logarithmic. So, this is basically 2 to the order n hat times some log log factors. Okay, and n hat is the n over root log n.
- u153 Ja.Yeah.
- u154 >> Ali je prav razmišljati, da je računanje vseh teh odvodov nekako lažje kot povečanje L, in če je, zakaj?>> Uh is it correct to think that like computing all these derivatives is somehow easier than making the L larger, and if so, why?
- u155 >> Odlično. Ja. Lahko bi rekli: »Zakaj smo to tako zastavili? Morda bi lahko dobili precej večji L in bi morali izračunati manj odvodov. To bi bilo dobro.«>> Great. Yeah. Um you might say like, "Why did we set it up like this? Like perhaps we could have gotten much bigger L and then we have to compute fewer derivatives. That would be uh good.
- u156 Odlično vprašanje. Bistvo je, mislim, kot bomo upam videli ali začeli spoznavati, da je že to nekako težko doseči. Morda je najtežje sploh dobiti velik L. In pri našem pristopu nikoli ne boste mogli dobiti L, večjega od log n.Great question. I guess the point is as we'll hopefully see or get into seeing um it's sort of already hard to do this. Like maybe the hardest thing is to get L large at all. And in the way we're going to do it you'll never be able to get L bigger than log n.
- u157 Torej bomo morda kmalu videli motivacijo. Pri tem celotnem pristopu nikoli ne boste dobili L, večjega od približno log n. To morate torej preprosto sprejeti, potem pa si rečete: moja stopnja je tako velika, dobim pa lahko samo L, log n.So this is a maybe we'll see the motivation of it uh uh soon. So this whole play you're never going to be able to get L bigger than like log n. So you have to like just suffer that and so then you're like well my degree is so large and I can only get like L um log n.
- u158 Takrat najprej pomislite: morda lahko dobim odvode, da bom imel dovolj podatkov, in to se izide, potem pa upajmo, da odvodov ni pretežko izračunati, in tudi to se ravno še izide. Pravzaprav bomo to morda poskusili motivirati s primerom, kjer je L dobesedno ena.That's where at first you're like well maybe I can get the derivatives so that I'll have like enough data uh and that works out and then hopefully the derivatives are not too hard to compute and that also just barely works out. In fact we're going to like maybe try to motivate this by like uh considering a situation where L is literally one.
- u159 Imeli bomo samo A0, težko stvar, ki jo poskušamo izračunati, in sestavili A1, ki ga je zelo lahko izračunati. Izra- izračunljiv je v času približno dve na koren iz n, kar je precej bolje. V nekem smislu bomo med njima lahko interpolirali, ampak L enako ena je tako malo, da boste potrebovali v bistvu vse odvode in ne boste dobili nobene pohitritve.So we'll just have A0 which is the hard thing we're going to try to compute and we'll make up an A1 which is like very easy to compute. It's com- computable in time like two to the root n which is much better. And I guess in some sense we'll be able to interpolate between them but then L equals one is like so low that you'll need basically all the derivatives and then like you won't get any speed up at all.
- u160 Lepota je torej morda v tem, da bomo lahko L nekoliko povečali, da bo število odvodov nekoliko manjše od n. In [prhne] mislim, da se bo pokazalo še nekaj, ko bom morda govoril o primeru L enako ena.So maybe the beauty is that we'll be able to make L a little bit large so that the number of derivatives can be like a little bit less than n. And [snorts] I guess like another thing that will come up is like when I maybe I talk about the L equals one case.
- u161 Kot sem rekel, te permanente bomo lahko izračunali v času približno dve na koren iz n, in to je skoraj predobro. Veste, da niste na pravi poti, ker je dve na koren iz n prehitro. Uporabili bomo dejstvo, da nas ne bi motilo, če bi lahko permanent A I izračunali zgolj v času dve na N skozi polilog N.As I said we could we'll be able to compute these permanents in time like two to the root n and that's like almost like two good. It's like you you know you're not on the right track because like two to the root n is like too fast. Like we're going to like use the fact that like we We mind if we could merely compute the permanent of the AIs in time like two to the N over poly log N.
- u162 To bi bilo povsem v redu. Če si dovolimo to šibkost, bomo lahko dobili bolj strukturirane A I, ki jih lahko uporabimo za naš interpolacijski načrt. Prav. Kakšno vprašanje? Prav.That would be perfectly fine. So, by allowing ourselves this like weakness, we'll be able to like get more structured AIs that we can use for our interpolation plan. Okay. Any questions? Okay.
- u163 Prav, to je dobra točka za premor. Naj poskusim ugotoviti, kaj povedati naprej. Ja.Uh okay, this is a good this is like a good pausing point. So, let me try to figure out what to say next. Yeah.
- u164 >> Rad bi postavil zelo podobno vprašanje. Rekli ste, da je bila ta tako imenovana relacija dovoljenja že znana v teoretičnem računalništvu.>> I I want to make a question very similar to this like so you said that this like so-called permission relation was was like already known in TCS.
- u165 >> Ja.>> Yeah.
- u166 >> Bolj na splošno me zanima, zakaj bi bilo pridobivanje podatkov iz odvodov lahko kaj lažje od pridobivanja iz drugih točk ničtega reda>> Uh I'm curious I guess more generally of why is it getting data points out of the derivatives can be is any easier than getting it out of other points of like the zero order
- u167 >> Pridobivanja na različnih mestih?>> Uh getting it other than like at like different locations?
- u168 >> Da.>> Yes.
- u169 >> Ja, dobro vprašanje. Zgodilo se bo nekako to, da bomo zgradili matrike A I, ki so na neki način podobne A, ampak imajo veliko dobre strukture, in uspelo nam jih bo narediti samo logaritemsko mnogo. Zato želimo iz teh krasnih strukturiranih A I iztisniti čim več.>> Um Yeah, that's a good question. Like somehow what's going to happen is we're going to build these matrices AI that are somehow like similar to A but have like a lot of good structure and we're somehow only going to be able to manage to like make logarithmically many of them. So, we want to like just like squeeze as much juice out of these uh great structured AIs as we can.
- u170 In to je tukaj motivacija, zakaj rečemo: potrebujem jih več. Zato bom poleg permanentov poskusil izračunati še te odvode. Morda bom zdaj omenil, ker se ne bomo spuščali v podrobnosti, naslednje.And I guess that's here the motivation why we're like, well, I need more of them. So, besides computing the permanence, I'll try to compute these derivatives. Um Uh maybe I'll mention now because um we're not going to get into details. Uh I'll mention the following.
- u171 A I bo zelo strukturiran, kot bomo videli čez trenutek, zato lahko njegov permanent učinkovito izračunamo. Kako izračunamo odvode permanenta na teh različnih me- strukturiranih mestih A I?Um AI uh is going to be very structured as we'll see in a second, which is why it's able to compute the permanence uh efficiently. Um How do you compute the derivatives of the permanent at these various lo- structured locations AIs?
- u172 Prav, to je nekakšna domača naloga za nekoga, ki je dober v algebraični zahtevnosti ali algebri permanentov ali pozna stvari, recimo zna zapisati formulo za permanent vsote dveh matrik.Um Okay, this is like kind of homework for a person who's like good at algebraic complexity or algebra permanence or knows things like who could write down like the formula for the permanent of the sum of two matrices.
- u173 Nekako vidite, da za izračun K-tega odvoda permanenta pri A I potrebujete, zadostuje, da znate izračunati permanente množice podmatrik A I.Um you kind of see that like to compute like the Kth derivative um of the permanent at AI, you kind of need to it suffices for you to be able to compute the permanence of a bunch of submatrices of AI.
- u174 In to je nekakšna preprosta lastnost permanenta. Zato sem to dodal. Hotel sem opozoriti, da A I ni samo dovolj strukturiran, da hitro izračunate njegov permanent, ampak lahko hitro izračunate tudi permanent katere koli podmatrike. Za izračun K-tega odvoda morate, če pogledate podrobnosti, izračunati nekako vse permanente podmatrik K krat K ali nekaj takega, morda glavnih podmatrik tega A I.Um and that's just like some sort of simple property of permanent. That's why I threw this in. So like I I sort of wanted to recall that like not only is AI like structured enough that you compute its permanent fast, you can also compute all any submatrix's permanent fast. So to compute the derivative like the Kth derivative, you have to compute uh you look into it and you have to compute like sort of all the K by K uh or something like all the K by K submatrix uh um maybe the principal ones uh permanence of this AI.
- u175 In tukaj to pride prav. Torej vse te lahko hitro izračunamo zaradi tega. Koliko pa je glavnih podmatrik K krat K? Približno N nad K, zato je tukaj ta N nad K. To je vse, kar bom v nadaljevanju povedal o računanju odvodov.And that's where this comes in. So like basically and these are all computable fast because of this. And then like how many K by K submatrices are there of principal ones is like N choose K, which is why this N choose K is here. So uh this is all I will say about this derivative computation stuff uh for the rest of this.
- u176 Ja.Yeah.
- u177 >> Če lahko dobite L točk, zakaj potem to pomeni, da lahko K postane nič?>> So if you're able to get the L points, why does it mean K can get zero though?
- u178 >> Zato ker imajo pri tej postavitvi, kot bomo videli, A1 do AL zelo prijetno strukturo. Povedal bom čez trenutek. Skoraj bločno diagonalne so in zato je njihove permanente zelo lahko izračunati. A0 pa je samo prvotna grozna, grozna nestrukturirana matrika.>> Uh it's because in this the setup like as we'll see A1 through AL have some like very enjoyable structure. I'll say it in a second. They're like kind of block diagonal almost and it makes them really easy to compute their permanence. But A0 is just like the original horrible horrible unstructured matrix.
- u179 Zato je sploh ne znate učinkovito izračunati. Ja. Celotna igra je torej narediti A1 do AL dovolj strukturirane, da jih lahko učinkovito izračunate, v podeksponentnem času, in hkrati tako, da z interpolacijo skozi njih dobite permanent A0, ki je zelo nestrukturirana matrika.So you don't know how to compute it efficiently at all. Um yeah. So this is like a whole game like make A1 through AL they're like structured enough that you can compute them efficiently like subexponential time and moreover in such a way that like by interpolating through them you can get the permanent of A0, which is this like very unstructured matrix.
- u180 Prav, pogovorimo se o strukturi, ki jo bodo imeli A I in zaradi katere jih bo lahko izračunati. Zakaj bodo A I nekoliko lažji?Okay, so let's talk about the kind of structure that AIs are going to have that will make it easy to compute them. Okay, so why will the AIs be a little bit easy?
- u181 >> [aplavz]>> [applause]
- u182 >> To seveda za I večji od nič. No, I in L. Prav. Na grobo skoraj zato, ker bodo videti takole. Bločno diagonalni bodo. Prav, takole bodo bločno diagonalni.>> This is of course I greater than zero. Well, I and L. Okay. At a high level, it's almost because they're going to look like this. They're going to be block diagonal. Um Okay, they're going to be block diagonal like this.
- u183 Prav, malo lažem. Laž bom pojasnil čez trenutek. Naj to matriko poimenujem B. A I ne bo čisto takšen kot B, ampak naj povem to. Poleg tega bodo vse te dimenzije približno n skozi, tukaj lahko postavite kateri koli polilog.Okay, I'm lying a little bit. I'll explain the line in a second. Um let me call this this matrix B. AI is not going to quite look like this B, but let me say this. Um and moreover, uh all of these dimensions will be like n over You can put any poly log here.
- u184 Morda torej n skozi log na deseto n. Aha, povedati moram, da bodo imele to poli- to bločno diagonalno strukturo, ko ustrezno preuredite vrstice in stolpce. Pravzaprav vas lahko vprašam.So maybe n over log 10 of n. Okay, so uh oh, I should say they will have this poly this block diagonal structure once you appropriately rearrange the rows and columns. So actually, maybe I can ask you.
- u185 Zakaj ste zelo veseli, ko morate računati permanent teh matrik? Matrik, ki so videti takole?I mean, uh why are you very happy about computing the permanent of these matrices? Matrices that look like this?
- u186 >> Ker računanje permanenta gre pri n skozi polilog, kar je dve na n skozi polilog.>> Because the computer like permanent goes at n over poly log, which is two to the n over poly log.
- u187 >> Ja, permanent bločno diagonalne matrike, kakršna je ta, je tako kot determinanta, ali pa se spomnite, kako permanent deluje. Permanent celote je permanent tega bloka krat permanent tega bloka krat permanent tega bloka krat permanent tega bloka in tako naprej. In koliko časa potrebujemo za permanent tega bloka?>> Yeah, so the permanent of like a block diagonal matrix that looks like this um it's just like determinant or you can remember from how permanent works. The permanent of this whole thing is just like the permanent of this block times the permanent of this block times the permanent of this block times the permanent of this block and so forth. And how long does it take to compute the permanent of this block?
- u188 Uporabimo to. Torej dve na dimenzijo, ki je, kar koli, n skozi polilog n. Vse te lahko torej izračunamo v tem zelo »hitrem« času.Well, we just use this. So, it's uh it's two to the dimension, which is like whatever, n over poly log n. Okay, so we can compute all of these in this like very {quote} fast time.
- u189 In potem, ne vem, manj jih je od tega. Mislim, da jih je logaritemsko, polilogaritemsko mnogo. Rezultate samo zmnožimo. Če bi bil A I tako čudovito strukturiran na ta bločni način, bi to vsekakor imeli. Ne bo čisto tak, bo pa blizu.Uh and then I don't know there's less than it. I guess there's logarithmically poly logarithmically many of them. We just multiply the results. So, if AI was like so glorious as to be structured in this block structured way, uh we would definitely have this. Uh it's not going to be quite like that, but it's going to be close.
- u190 A I bo oblike takšnega bloka plus matrika napake. Takole, in ta matrika napake bo imela rang največ n skozi neki velik polilog n.AI is going to be of the form like block like this plus an error matrix. Like this, and this error matrix is going to have rank at most n over some big poly log n.
- u191 Vprašal vas bom torej: »Predstavljajte si, da bločno diagonalnega dela ni in je A I samo ta matrika nizkega ranga. Zakaj bi bili veseli, da morate izračunati njen permanent?« Prav, dobro.Um So, I'll ask you like, "Okay, now imagine that like the block diagonal part wasn't here, and you just had that AI was this low rank matrix. Why would you be happy about computing its permanent?" Okay, good.
- u192 Tudi jaz nisem vedel, ampak vsak, ki kaj ve o algebraični zahtevnosti ali podobnih stvareh, bi vedel, da je to staro in ne prav težko dejstvo. Morda ga je prvi dokazal Barvinok, ampak mislim, da ni zelo težko: tudi permanent matrik nizkega ranga lahko učinkovito izračunate.I also didn't know, but everybody who knows anything about algebraic complexity or this kind of stuff would know that it's like an ancient and like not very hard fact. Maybe it was first proved by Barvinok, but I guess it's not very hard, that you can also compute the permanent of low rank matrices efficiently.
- u193 Torej lahko poleg tega dejstva tudi rang R obdelate v času N na red R. Pravzaprav lahko, če ste nekoliko pazljivi, to naredite v času N plus R nad R.Uh so, maybe in addition to this this fact, you can also do rank R in time N to the order R. Actually, in particular if like you're a little bit careful, you can do it in time N uh plus R choose R.
- u194 In na ta način to nekako zajema tudi tisto, ker je največji možni rang N, potem pa dobite 2N nad N, kar je približno 4 na N. Torej je pravzaprav slabše. To je dve na N, to pa štiri na N. Ampak pri tem algoritmu nas pri eksponentnih izrazih konstantni faktor tukaj zgoraj ne moti.And it's In that way, like this kind of subsumes this because the worst the rank could be is N and then you'll get like 2N choose N, which is around 4 to the N. So, actually it's worse. It's This is like 2 to the N, this is 4 to the N. But, actually in this this uh this algorithm, whenever we're doing like exponential, like we don't mind constant factor up here.
- u195 Zato je tudi to v redu. Ja, to je bila stranska pripomba. Ni pomembno. Lahko bi vprašali: »Kako pa to naredite?« In o tem vam nočem govoriti, ker je to staro, dobro znano dejstvo. Ampak povedal bom, da lahko to naredite tudi z dinamičnim programiranjem. In povedal bom celo droben, naj vam samo zastavim to kot majhno vajo.So, this is also fine. Yeah, that was like a like a side point. Never mind. Um So, you might say, "Okay, how do you How do you do this?" And I don't want to tell you about it because I'm like, "Oh, this is like an ancient and well-known fact." But, I will say like, "Okay, you can also do it with dynamic programming." Um and I'll even say like a tiny Let me just set it up for you as a little exercise.
- u196 Kaj pomeni, da ima A rang, rang A je enak R? Pomeni samo, da lahko A zapišemo kot vsoto R matrik ranga ena, ki so zunanji produkti.Um What does it mean for A to have you know, rank A equals R? Uh it just means that A can be written as like uh the sum of well, R rank one matrices, which are like outer products.
- u197 Če se spomnite linearne algebre, je videti takole. Ali če je vse kombinatorično, ničle in enice ter dvodelna prirejanja, prav, potem pa pustimo to. Ampak prav.If you remember your linear algebra, it looks like this. Or if everything is like combinatorial and zeros and ones and bipartite matchings, Okay, and then never mind about that. Um But, okay.
- u198 Če imate matriko ranga ena, to samo pomeni, da je element IJ enak UI krat VJ. In to je res lahka vaja. Zagotavljam vam, da lahko pri takšni matriki, če poznate U in V, njen permanent izračunate povsem trivialno.If you have like rank one matrix, uh just means that like the IJ entry is like UI times VJ. And this is a truly easy exercise. I assure you like a matrix that looks like this, you know, if you know the U and V, you can compute its permanent extremely trivially.
- u199 In potem, v redu, če je v splošnem videti kot plus UI prime VI prime, je to primer ranga dve. Zdaj morate že malo razmišljati in uporabiti svoje veščine dinamičnega programiranja, ampak bi šlo. Torej si bomo zapomnili tudi to dejstvo.And then, okay, if it in general looks like plus UI prime VI prime, that's like the rank two case. Okay, now you have to start thinking a little bit and using your dynamic programming skills, but you could do it. Okay, so we're also going to like save up this fact.
- u200 Ja.Yep.
- u201 >> Ali mora biti ta polilog večji od ena, mislim eksponent? Sicer pa>> Do we need like the poly log to be greater than one, like the exponent? Or otherwise
- u202 >> Aja, ja, ja, ja. Kadar tukaj na tem predavanju pišem polilog, mislim logaritem na neko veliko potenco, ki jo lahko sami določite. Če torej potenca logaritma za vas ni dovolj visoka, lahko prilagodite kon- konstante drugje, da bo tako velika, kot želite.>> Oh, yeah, yeah, yeah. Whenever I'm writing uh here in this this lecture poly log, I mean like log onto a big number, which you yourself can control. So, if it's not like not a high enough power of log for you, you can adjust the con- constants elsewhere to make it as big as you want.
- u203 Torej lahko morda to naredimo ... Pravzaprav, v redu, za tega bom celo rekel, da uredite tako, da je log na 100 od n. To je tisti polilog, za katerega želite, da je največji. Em, v redu.So, maybe we can make this In fact, okay, like this one I will even say you arrange for it to be like log to the 100 n. This is the one of the poly log that you want the biggest. Uh okay.
- u204 Torej, rekel sem, da če bi bil AI tako prijazen, da bi bil videti kot ta bločna matrika, bi lahko njegovo permanento izračunali zelo učinkovito. Če pa bi bil AI tako prijazen, da bi bil videti kot ta napaka, kot matrika nizkega ranga, bi jo lahko izračunali učinkovito, ker ... No, tega nisem povedal, ampak to bi naredili v času n na potenco reda n deljeno z log na 100 od n.So, actually I said that like if AI was so nice as to look like this block matrix thing, you could compute its permanent very efficiently. If on the other hand AI was so nice as to look as like this error like low rank matrix, you could compute it efficiently because Okay, I didn't say it, but you would do it in time n to the power of order n over log 100 to the n.
- u205 To je torej dva na potenco reda n deljeno z log na 99 od n, kar je še vedno odlično. Ampak dejansko bo te oblike, B plus E.So, that's two to the order n over log to the 99 n, but that's still great. Uh but actually it's going to be of this form, B plus E.
- u206 >> [se odkašlja]>> [clears throat]
- u207 >> Zdaj se torej vprašate: aha, ali lahko to združimo? Če imam bločno diagonalno matriko z, saj veste, nekoliko majhnimi bloki plus matriko nizkega ranga. Ali lahko tudi to izračunam v času približno dva na potenco n deljeno s polilogom? Da.>> So, now you're like, oh, well, can we put this together? Like if I have a matrix that's block diagonal with like, you know, slightly small blocks plus a low rank matrix. Can I also compute it in time like two to the end over polylog? Yes.
- u208 To je še ena vaja. Vse, kar ste naredili tukaj, je bilo z dinamičnim programiranjem, in če torej vse združite, lahko, če je velikost bloka S in rang R, s povezovanjem teh idej izračunate permanento B plus E v času približno dva na potenco reda S krat N na potenco reda R.It's like one more exercise. Like, okay, everything you did here was like by dynamic programming and so if you like put it all together, you know, you can compute like if this block size is S and this rank is R uh by gluing these ideas together, you can compute the permanent of B plus E in time it's like two to the order S times N to the order R.
- u209 V redu? To bo torej naša končna predstava lepo strukturirane matrike. Bločno diagonalna z ne prevelikimi bloki plus napaka nizkega ranga. Te stvari so ravno dovolj strukturirane, da lahko njihove permanente izračunamo res hitro.Okay? So, like this will be our like final notion of like nicely structured. It's like uh block diagonal with not too big blocks plus low rank error. And these things are just structured enough that you can compute their permanents real fast.
- u210 Res hitro pomeni dva na potenco n deljeno z velikim polilogom. Em, v redu. Še manjša opomba, in ne spomnim se, ali je pomembna ali ne, ampak drži.Real fast being two to the end over a big polylog. Um Okay. As a minor note, and I cannot remember if this is important or not, but it is true.
- u211 Torej, kaj se tukaj v resnici dogaja? Naj to pobrišem. Oh. V redu. Da izpolnimo vse obveznosti, naj bi znali to narediti tudi za poljubno podmatriko AI. Ampak to je pravzaprav čisto v redu, ker je razred bločno diagonalnih matrik plus matrik nizkega ranga zaprt za jemanje podmatrik.So, uh what's really going on here? Okay, let me erase this. Oh. Okay. To discharge all the obligations, we're also supposed to be able to do this for any submatrix of AI. But, that's actually quite fine because if you're block diagonal plus low rank, these are like closed under like submatrices.
- u212 Če vzamete podmatrike, je struktura še vedno bločna, rang pa se lahko samo zmanjša. Zato velja tudi ta stvar s podmatrikami. V redu. Pravzaprav naj rečem kar to.If you take submatrices, it's still like block structured and it's can only decrease the rank. So, that's why this submatrices thing is also true. Okay. So, in fact, let me just say this.
- u213 Em, kar je, predvidevam, preprosto to, da je za I, ki ni nič, AI lepo strukturiran kot, no, v tej obliki BI plus E.Uh which I guess is just that like four I is like not zero. AI is like structured nicely as um well, of this like form BI plus E.
- u214 V redu, kjer je rang E največ N deljeno s polilogom in velikost blokov B prav tako največ N deljeno s polilogom. In ta malenkost, za katero se ne spomnim, ali je pomembna ali ne, je, da je E dejansko isti za vsak I.Okay, where you know, rank of E is at most N over poly log and like block size of B is also at most N over poly log. And this minor thing, which I forget if it's important or not, is that actually it's the same E for every I.
- u215 Mislim, mislim, da ni pomembno. V redu, to je torej v nekem smislu novi načrt. Ima kdo na tej točki kakšno vprašanje? V redu.I I don't think it matters. Okay, so in some sense like this is the new plan. Okay, any questions at this point? Okay.
- u216 Tukaj se zelo rad ustavim in kar končam, ker si mislim: »V redu, to je kul načrt.« Naslednji del pa je: »V redu, ugotoviti moramo, kako narediti polinomsko družino matrik, ki se pri nič ujema z nestrukturirano A, pri preostalih pa ima to lepo, lepo strukturo.« Po eni strani se tukaj pojavi tista kul ideja, če sploh je kakšna kul ideja.This is where I like love to pause and just stop it because I'm like, "Okay, this is a cool plan." And now the next part is like, "Okay, we got to like figure out how to make like a polynomial family of matrices that like matches the unstructured A at zero, but has this like nice nice structure on uh, the remaining ones." And on one hand, this is where like the cool idea, if there's any cool idea, like comes in.
- u217 Po drugi strani pa je postavitev precej kočljiva in to je del, ki ga imam večinoma v glavi, ampak bojim se, da se bom zataknil in kaj zamočil, ampak to bomo zdaj naredili.On the other hand, setting up is like kind of finicky and it's like the point that I have like mostly in my head, but I like I fear I'm going to like get stuck in mess it up, but um, that's what we're going to do.
- u218 >> [prhne]>> [snorts]
- u219 >> Em, v redu. Kakšno vprašanje? Pravzaprav, ker me je strah, saj veste, da bom kaj zamočil in zabredel v zmešnjavo, v kateri se bom zmedel, naj nekoliko brez motivacije navedem še eno, no, morda ključno bistveno sestavino, ki nam pri tem pomaga.>> Uh, Okay. Any questions? Okay, actually, because I have this like fear of, you know, messing it up and getting into a morass where I get confused, let me somewhat unmotivatedly state uh, another well, maybe like the key essential ingredient that helps us do this.
- u220 Tako boste vsaj videli to lepo trditev, ki jo razumem. Potem pa bomo videli, ali lahko govorimo o tem, kako vse skupaj sestaviti. Em.And then at least you'll see this like nice statement, which I can understand. And then we'll see if we can talk about putting them together. Um
- u221 >> [prhne]>> [snorts]
- u222 >> Ja. V redu. Kaj je torej ta druga lepa trditev? Samo v mislih poskušam skicirati način, kako bi to povedal.>> Yeah. Okay. So, what is this other nice statement? Uh I'm just trying to sketch in my mind like a way to try to say this.
- u223 Em, v redu. Ključna osrednja, osrednja sestavina dokaza je nekaj, kar bi v analizi Boolovih funkcij ali teoriji zahtevnosti veljalo za splošno znano dejstvo.Um Okay. The key central central ingredient for proving this is what would be like a folklore fact in analysis of Boolean functions or complexity theory.
- u224 Pravzaprav smo v bistvu že videli natanko to dejstvo. Nisem ga izrecno navedel v tej obliki, ampak bilo je natanko tisto ključno dejstvo v ozadju, ki smo ga uporabili v dokazu Razborova in Smolenskega, da AC-ji, da, no, pri nas, da večinska funkcija zahteva vezja konstantne globine eksponentne velikosti.In fact, we've seen basically this exact fact before. I didn't explicitly state it in this way, but it was exactly the key underlying fact that we used in the Razborov-Smolensky proof that AC's that uh well, for us that majority requires exponential size constant depth circuits.
- u225 Em, ko smo dokazovali, da večinska funkcija ne more biti ... V redu, naj preskočim, kaj smo govorili, ampak to je bila ključna, em, ključna sestavina tistega dela. Gre pa za računanje FUNKCIJE PARNOSTI NA, EM, L bitih s polinomom nizke STOPNJE.Um when we were showing that majority cannot be uh Okay, let me skip what we're saying, but like it was the key um it was a key ingredient in that bit. And it's about computing uh PARITY FUNCTION ON UH L bits with a low DEGREE POLYNOMIAL.
- u226 V REDU? ZELO ME OBREMENJUJE, DA ne rečem parnost na N bitih. Ker ko delaš, kaj vem, analizo Boolovih funkcij, bi moralo biti število bitov N, ampak je prepozno. N je ta n, zato moramo ... to je ta l. Em, v redu.OKAY? IT'S STRESSING ME OUT A LOT TO not say that it's parity on like N bits. Uh cuz like when you're doing I don't know, analysis of Boolean functions, the number of bits should be N, but it's too late. N is this n, so we have to like it's this l. Um Okay.
- u227 In upam, da boste med tem začeli opažati nekakšno ujemanje vzorcev med tem in tem. Potem pa moramo to ujemanje vzorcev še dopolniti. Em, v redu, naj malo podrobneje povem, kaj s tem mislim. Pravzaprav smo govorili o ... V redu, ni važno.And hopefully as I'm doing this you'll start to see some kind of like pattern matching a little bit between this and this. Which then we have to fill in that pattern matching. Uh okay, so let me say a little bit more what I mean by this. Um So actually we've talked about Okay, never mind.
- u228 Torej bomo pri teh stvareh bite predstavljali s plus ali minus ena namesto z nič in ena. V REDU, TO JE ČISTO V REDU. TO JE SAMO DOGOVOR. IN TE SI PREDSTAVLJAMO KOT realna ali racionalna števila.So um for this stuff we're going to represent bits by plus or minus one instead of zero one. OKAY, THAT'S FINE. IT'S JUST A CONVENTION. AND LIKE THESE ARE THOUGHT OF AS like real numbers or rational numbers.
- u229 V redu, potem imamo funkcijo parnosti na n bitih, l bitih. Vidite, že mešam. Em, x ena do x l. In to lahko izrazimo kot zelo preprost realni polinom več spremenljivk v x ena do x n, kadar uporabljamo zapis, v katerem plus ali minus ena pomenita resnično in neresnično.Okay, and then we have the parity function on n bits, l bits. See, I'm already messing it up. Uh x one through x l. And this can be expressed as a very simple real multivariate polynomial in x one through x n when we use the notation that plus or minus one is true and false.
- u230 Bi mi kdo povedal, kakšen je? Ja, preprosto njihov produkt.Uh does somebody want to tell it to me? Yeah, it's just the product of them.
- u231 >> [aplavz]>> [applause]
- u232 >> To samo kaže, kako pomembno je, nad katerim poljem delate. Če na primer bite predstavljate z nič in ena ter delate nad poljem po modulu dve, ima parnost polinom stopnje ena. Preprosto seštejete bite po modulu dve. Toda v tem svetu, kjer uporabljate plus ali minus ena kot realna števila in poskušate izračunati parnost, ima parnost na l bitih polinom največje možne stopnje, x ena, produkt x ena do x l.>> So it just goes to show it makes a lot of difference like what um field you're working with. For example, if you're representing your bits as zero and one and you're working with a field mod two, then parity has a degree one polynomial. You just sum the bits mod two. But in this world where you're using plus or minus one as real numbers, trying to compute parity, uh parity on l bits has like a maximal degree polynomial x one product of x one through x l.
- u233 Em, pravzaprav vsaka ... Mislim, to je preprosto dejstvo. Mislim, da smo o tem že govorili, ampak vsaka funkcija, ki preslika n bitov v realna števila, ima polinomsko predstavitev stopnje največ L.Um actually every I mean this is a simple fact. I I think we talked about it before, but like every function mapping n bits into real numbers uh has a polynomial representation of degree at most L.
- u234 Pravzaprav zato, ker jo lahko interpolirate z nekim polinomom in nikoli ne potrebujete xi na kvadrat ali xi na kub ali xi na četrto, ker če vas zanimajo samo vhodi plus ali minus ena, je xi na kvadrat preprosto ena. Torej mora biti polinom samo multilinearen, zato nikoli ne bo imel stopnje več kot L. Parnost pa potrebuje največjo stopnjo L.In fact, because you can interpolate it by some polynomial, and you never need to use xi squared or xi cubed or xi to the fourth because if you only care about inputs which are plus or minus one, then like xi squared is just one. So, it only needs to be a multilinear polynomial, so it'll never have degree more than L. Parity, though, needs maximal degree L.
- u235 Je nekako najbolj zapletena funkcija z vidika stopnje. Toda ključna ideja, ki jo želimo uporabiti, pravi, da jo lahko nekako izračunamo s polinomom precej nižje stopnje.It's like the It's like kind of like the most complicated function from the terms of degree. But, uh this key idea we want to use is saying that oh, somehow we can compute it um with a much lower degree polynomial.
- u236 No, izračunali jo bomo približno. Em, v redu. In, em ...Well, we're going to approximately compute it. Um okay. And uh
- u237 >> [prhne]>> [snorts]
- u238 >> Drugo vprašanje je, v kakšnem smislu jo bomo približno izračunali. Ker obstajajo različni pojmi tega, da en polinom približno izračuna drugega. Torej bomo iskali, recimo, polinom, em, polinom n spremenljivk, L spremenljivk, recimo G od X1 do XL nizke stopnje D, in ta D bo ta D.>> Another question is like, what is the sense in which we're going to approximately compute it? Because there's different notions of you know, one polynomial approximately computing another one. Um so, we are going to seek let's say a polynomial, um n-variate polynomial, L-variate, let's say G of X1 through XL of low degree D, and that D is going to be this D.
- u239 Em, pravzaprav sem pozabil, kaj se tukaj dogaja. Morda je takole. Bistvo je, da se spomnite L, v redu, tukaj ni zapisan, ampak L je bil, če se spomnite, nekaj takega kot log n. Torej je D nekaj takega kot kvadratni koren L.Um I forget what's going on here, actually. Maybe it's like this. Um so, uh the point is that remember L, okay, it's not written here, but L, if you recall, was like something like log n. So, this is like D is like something like square root L.
- u240 In to je vir našega prihranka, dejstvo, da ... oziroma moral bi reči, da je vir prihranka to, da nam uspe narediti ta D opazno manjši od L. Nekako bi bilo trivialno, če bi bil ta D lahko L, ampak mora biti manjši od L, in dejstvo, da ga zmanjšamo za faktor koren iz log n, je tisto, od koder izvira ves prihranek.And this is like the source of our savings, the fact that like or I should say the source of our savings is that this D manages to be like noticeably less than L. And somehow like it would be trivial if this D could be L, but we need it to be smaller than L, and the fact that we get it smaller by a factor of root log n is like where the whole savings comes from.
- u241 To se tukaj dogaja. Torej bo ta D, saj veste, 100 koren L log L, kar je, em, recimo epsilon blizu parnosti.That's going on here. So, this D will be, you know, 100 root L log L that is um let's say epsilon close to parity.
- u242 Em, in povedal bom, kaj [prhne] s tem mislim. Najprej, epsilon bo 1 deljeno z L na 100. Torej zelo majhen, kot 1 deljeno z ... V redu, ne vem, ali so te stotice enake, ampak ni važno. Em, blizu so.Um and I'll say what [snorts] I mean by this. So, first of all, epsilon is going to be 1 over L to the 100. So, very small, like 1 over Okay, I don't know if these 100s are the same, but never mind. Uh they're close.
- u243 Zelo majhen, kot 1 deljeno z inverznim polinomom. To pa pomeni, da je verjetnost po naključnem X, da se G od X razlikuje od parnosti L od X, največ epsilon.Very small, like 1 over inverse polynomial. And what this means is uh the probability over random X that G of X differs from parity L of X is at most epsilon.
- u244 To je nekakšen pojem bližine v smislu L0 ali Hammingove razdalje. Torej bo ta G na večini vhodnih nizov povsem pravilen. Na deležu 1 minus epsilon.So, it's kind of like a L0 or like Hamming distance notion of uh closeness. So, this G is going to be like exactly right on most input strings. 1 minus epsilon fraction.
- u245 >> [prhne]>> [snorts]
- u246 >> Na preostalem deležu epsilon pa je lahko povsem mimo. Lahko je neka nora številka, velikanska številka. Morda ni plus ali minus ena. Bo neko čudaško racionalno število. Ampak to je v redu. In to [prhne] bo vir ... Ta napaka bo vir te zgodbe z rangom.>> And then on the remaining epsilon fraction, it can be like totally off. It can be like a crazy number, a giant number. It might not be plus or minus one. It'll be some weirdo rational number. Um but that's fine. Uh And this [snorts] is going to be the source This error is going to be the source of this rank business.
- u247 Em, [prhne] ne morem ... Težko se spustim v to. Čas imam omejen, ampak na visoki ravni: če pogledate resničnostno tabelo G minus parnost L, ima samo epsilon krat dva na L neničelnih vrednosti.Um [snorts] I can't It's hard for me to get into it. I only have so much time, but at a high level, uh, if you look at the truth table of G minus parity L, uh, it only has epsilon times two to the L non-zeros.
- u248 Pravim resničnostna tabela, čeprav pravzaprav od funkcije s poljubnimi vrednostmi odštevam funkcijo z vrednostmi plus ali minus ena, ampak v tabeli vrednosti to pomeni natanko to, da so večinoma ničle, razen na epsilon krat dva na L mestih. Oziroma na takem deležu mest.Okay, I say truth table, it's like actually I'm subtracting a plus or minus one valued function from like an arbitrary valued function, but just in the table of values, um, this is exactly saying that like it's mostly zeros except for epsilon times two to the L spots. Or like a this fraction of spots.
- u249 In upam, da vam to malo diši po rangu. Nekako uporabljamo dejstvo, da če imate dia- ... To ne bo imelo smisla, ampak poskušam vam dati občutek. Če imate diagonalno matriko, kjer so večinoma ničle, tu in tam pa kakšna zvezdica, je rang te matrike enak številu zvezdic.And hopefully this has like a little the flavor of, um, rank for you. Uh, sort of we're using the fact that if you have like a di- this is not going to make sense, but I'm trying to give you the flavor. Like if you have like a diagonal matrix where like you have like mostly zeros, but uh, a few stars, then the rank of this matrix is like the number of stars.
- u250 In zvezdice bodo ustrezale tem mestom, kjer se resničnostni tabeli razlikujeta. To je samo za občutek. Za zdaj vam še ne more biti zares smiselno. Em, v redu. Bistvo pa je, da je delež napake oziroma relativni rang približno ena deljeno z L na neko veliko potenco, 100.And like stars are going to correspond to like these places where the truth table differ. This is just like some flavor. It cannot really make sense to you yet. Um, okay. And the point of this is that, uh, sort of the fractional wrongness or the fractional rank is like one over L to the like a big number 100.
- u251 In ne pozabite, L je nekako log N.And remember L is sort of like log N.
- u252 >> [prhne]>> [snorts]
- u253 >> Torej prav od tod izvira ta log N tukaj. To je v bistvu celo N log na 100 od N. V redu, [prhne] dejstvo, da je ta relativni rang približno ena deljeno z log na 100 od N, bo izviralo natanko iz tega, da je bil delež napake približno ena deljeno z L na 100.>> So this is exactly the source of the log N right here. Like this is even like N log to the 100 N basically. Okay, it's [snorts] like the fact that this fractional rank is like one over log to 100 N is going to be be exactly because this like fractional wrongness was like one over L to 100.
- u254 Ampak tega še ne morete videti, ker tudi ne bomo imeli časa zares pojasniti, kako se bo ta sestavina preslikala v to sestavino, ampak ... Lahko si preberete več ali pa, ne vem, morda bom naredil kak dodaten video, kjer dokončam skico, ali kaj takega, ampak v redu, vrnimo se k temu osnovnemu, kaj vem, problemu iz analize Boolovih funkcij.But you cannot see that yet cuz we're also not going to have time to really explain how this ingredient is going to map into this ingredient, but uh Oh, you can read more about it or I don't know if I'll like make some side video where I finish the sketch or what, but Okay, so let's go back to this basic I don't know analysis of Boolean functions problem.
- u255 In, saj veste, če ste pravi navdušenec nad teoretičnim računalništvom, morda že natanko veste, kako to rešiti, ampak vam bom pokazal. Em, v redu, torej, ja, če ste ... Vsi ti polinomi, ki nas zanimajo, imajo nekako stopnjo do L.And uh You know, if you're a real TCS enthusiast, like maybe you already exactly know how to solve this, but I'll I'll show you. Um Okay, so uh Yeah, if you are all these polynomials that we care about kind of degree up to L.
- u256 Parnost ima stopnjo natanko L, mi pa poskušamo njene vrednosti ujeti s polinomom precej nižje stopnje. Kakšna je torej ideja? Eden od vidikov, ki bo pomemben, je, da je parnost simetrična funkcija. Njena vrednost je odvisna samo od Hammingove teže vhoda.Parity has degree exactly L, but we're trying to like match its values with a much lower degree polynomial. Um so, what's the idea? Uh one aspect that's going to be important is that um parity is a symmetric function. Uh its value only depends on the Hamming weight of the input.
- u257 Drugače povedano, to se bo še pojavilo: parnost L od X ena do XL je odvisna samo od X ena plus tri pike plus XL.Another way to say that is this is going to come up. Uh parity L X one through XL only depends on X one plus dot dot dot plus XL.
- u258 To ni natanko Hammingova teža, ker uporabljamo zapis s plus ali minus ena. Če bi uporabljali zapis z nič in ena, bi to bila Hammingova teža. Ampak v redu, do nekega linearnega premika je enako kot Hammingova teža. Če seštejete pluse ali minuse ena, vam to pove Hammingovo težo.This is not exactly the Hamming weight because we're using plus or minus one notation. If we're using zero one notation, this would be the Hamming weight. But okay, it's it's the same as the Hamming weight up to some linear shift. Like if you add up uh plus or minus ones, like this tells you the Hamming weight.
- u259 >> [prhne]>> [snorts]
- u260 >> Em, in na splošno, če vas zanimajo simetrične funkcije na ... Tukaj bi moral biti L. Em, na plus ali minus ena na L.>> Um And in general if you're interested in symmetric functions on This should have been L. Uh plus or minus one to the L.
- u261 >> [prhne]>> [snorts]
- u262 >> Em, lahko jih izrazimo kot ... Tudi to je zelo preprosto. Lahko jih izrazimo kot>> Um they can be expressed as This is also very simple. Uh they can be expressed as
- u263 >> [aplavz]>> [applause]
- u264 >> polinom ene spremenljivke P v X ena plus tri pike plus XL. Seveda bo vsaka funkcija iz L-bitnih nizov v R, ki jo lahko izrazite kot polinom ene spremenljivke, uporabljen na vsoti X1 do XL, simetrična, ker se vsota ne spremeni, če permutirate XI-je.>> a univariate polynomial P in X one plus dot dot dot plus XL. Um Of course, any function from L-bit strings to R which you can express as a univariate polynomial applied to the sum of X1 through XL will be symmetric because if you permute the XIs, it doesn't change their sum.
- u265 Nekaj dela je treba, da vidimo obratno smer, da je torej vsako tako simetrično Boolovo funkcijo mogoče izraziti kot polinom ene spremenljivke, uporabljen na vsoti, ampak ni težko. To je povezano tudi s Trevorjevim videom, kjer je govoril o simetričnih funkcijah.Uh Takes a little bit of work to see the converse that conversely every symmetric Boolean function like this can be expressed as univariate applied to the sum, but it's not hard. Um it's also related to uh Trevor's video where he was talking about symmetric functions.
- u266 Ampak je v tem primeru malo lažje. In tudi stopnja tega polinoma P bo enaka stopnji vaše simetrične funkcije kot polinoma več spremenljivk. Ker tukaj vzamete nekaj stopnje ena in to vstavite v polinom P neke stopnje D, boste dobili polinom več spremenljivk stopnje D.Uh but a bit easier in this case. And uh also like the degree of this polynomial P will be the same as the multivariate degree of your symmetric function. Because here you're like taking a degree one thing and plugging it into like a polynomial P of some degree D, you'll get a like a degree D multivariate polynomial.
- u267 Torej to posebej pove, da lahko parnost tako izrazimo z nekim čudnim polinomom P stopnje L. In to je res precej čuden polinom. Videti je precej lepše, ko ga zapišete tako, ampak je mogoče. Poskusimo narisati sliko.So in particular says that like parity can be expressed like this for some funny degree L polynomial P. And it's sort of a a funny polynomial. So it looks much nicer when you write it like this, but it is possible. So let's try to draw a picture.
- u268 Em, tako nekateri radi rišejo Boolovo kocko, hiperkocko. To je slika 01 na L ali plus ali minus ena na L, pri kateri Hammingova teža narašča navzgor.Um This is how some people like to draw the Boolean cube, the hypercube. This is a picture of 01 to the L or plus or minus one to the L with like Hamming weight going up.
- u269 Torej je to Hammingova teža ... Sliko bom čez trenutek pojasnil, če je še niste videli. Hammingova teža L ali enakovredno vsota XI je enaka plus L. Morda sem ... Kakorkoli, ne zanemarjam, ali nič pomeni plus in ena minus ali obratno.So this is like Hamming weight I'll explain this picture in a second if you haven't seen it. Hamming weight L or equivalently sum of XI equals plus L. Maybe I'm whatever I'm not ignoring whether zero stands for plus and one stands for minus or vice versa.
- u270 To je Hammingova teža nič, kar je enakovredno temu, da je vsota xi enaka minus L. To je Hammingova teža L deljeno z dve, kar je enakovredno temu, da je vsota xi enaka nič, in tako naprej.Uh this is like Hamming weight zero, which is equivalent to sum of xi equals negative L. This is like Hamming weight L over two, which is equivalent to sum of xi equals zero, and so forth.
- u271 V redu, narisano je na ta čuden način, da vas spomni, da je med vsemi dva na L nizi tukaj en sam tukaj in en sam tukaj. Na naslednji ravni s Hammingovo težo L minus ena jih je zelo malo. Pravzaprav jih je L. In tudi tukaj jih je zelo malo pri ... Tukaj pri Hammingovi teži L minus dve jih je morda L nad dve.Okay, it's drawn in this like funny fashion to like remind you that like among all the two to the L strings here, there's like one here and like one here. And there's like very few at the next level Hamming weight L minus one. In fact, there's L. And there's very few here also at There's like maybe L choose two here at Hamming weight L minus two.
- u272 Ampak vse dogajanje je tukaj, ko ste okoli srednje Hammingove teže. Prav tukaj jih je natanko L nad L deljeno z dve, kar je zelo blizu dva na L deljeno s korenom L. In vse ravni, torej ta delež prav na sredini oziroma ta raven prav na sredini vsebuje delež ena deljeno s korenom L vseh nizov.But like all the action is like happening here when you're like around the middle Hamming weight. Right here, there's exactly L choose L over two, which is very close to two to the L divided by root L. And all the levels, so this fraction right in the middle or this level right in the middle contains a one over root L fraction of all strings.
- u273 In naslednja pri Hammingovi teži L deljeno z dve plus ena prav tako, če malo poračunate, vsebuje v bistvu delež ena deljeno s korenom L vseh nizov. Enako velja za vse tiste čisto blizu srednjih plasti. In posebej lahko dobite skoraj vse Boolove nize že, če greste do Hammingove teže L deljeno z dve plus ali minus koren L.And the next one at Hamming weight L over two plus one also, if you do a little calculation, contains basically a one over root L fraction of all strings. And like so do all the ones like really near the middle layers. And uh in particular, you can get like almost all the Boolean strings just by going out to Hamming weight L over two like plus or minus root L.
- u274 No, če greste do L deljeno z dve plus ali minus približno 50 koren L, boste dobili približno 99,999999 % vseh nizov. In če greste do L deljeno z dve plus ta količina, torej L deljeno z dve plus ali minus D, kjer je D koren L krat koren log L, potem s preprostim izračunom z uporabo, kaj vem, Chernoffove meje ali česa podobnega, dobimo dobro znano dejstvo, da te ravni vsebujejo delež 1 minus 1 deljeno z L na 100 VSEH NIZOV.Well, if you go like L over two plus or minus like 50 root L, you'll get like 99.999999 of percent of all the strings. And if you go out to L over two plus this amount, like L over two plus or minus D, where D is root L times root log L, then it's like a simple calculation using I don't know, Chernoff bound or whatever, well known fact that uh these levels give you like a 1 minus 1 over L to the 100 fraction OF ALL STRINGS.
- u275 V REDU, TOREJ SE SKORAJ VSE DOGAJANJE V BOOLOVI KOCKI dogaja znotraj te ravni. In kaj se dogaja s parnostjo? Ne pozabite, parnost je simetrična funkcija. Torej je vrednost parnosti pri poljubni Hammingovi teži oziroma je odvisna samo od Hammingove teže.OKAY, SO LIKE ALMOST ALL THE ACTION IN THE BOOLEAN CUBE is happening inside this level. And Okay, what's going on with parity? Remember parity is like a symmetric function. Uh so the value of parity on any Hamming weight or it only depends on the Hamming weight.
- u276 Parnost ima to lastnost, da je plus ena ... V redu, na srednji plasti ima neko vrednost plus ena ali minus ena, odvisno od parnosti L deljeno z 2. Potem je na naslednji plasti minus ena, na naslednji plus ena, minus ena, minus ena. Torej izmenjuje cele pasove od spodaj navzgor.So parity has this feature that like it's plus one Okay, it has some value plus one or minus one on the middle layer depending on the parity of L over 2. And then it's like minus one on the next layer and then it's like plus one on the next layer, minus one, minus one. So it like alternates the whole stripes from the bottom to the top.
- u277 In to je parnost. Em, naš cilj je torej dobiti to funkcijo G ali F, G, ne vem, zakaj sem to napisal. Ja, G, ki večinoma pravilno zadene parnost.And that's parity. Um And so this is our goal to get this function G or F, G uh I don't know why I wrote it. Yeah, G that mostly gets parity right.
- u278 In G naj bi imel nizko stopnjo. Sam G bomo vzeli simetričen. Torej bomo želeli, da je G ... Škoda. G bomo vzeli v obliki, em ...And uh G is supposed to have like low degree. Uh we're going to take G itself to be symmetric. Okay, so we're going to want G to be That's a shame. Uh we're going to take G to be of the form uh
- u279 >> [aplavz]>> [applause]
- u280 >> nekega polinoma ene spremenljivke Q, uporabljenega na X1 plus tri pike plus XL. In naš načrt bo preprosto ... V redu, zdaj bo ta polinom ene spremenljivke Q videti takole.>> like some univariate polynomial Q applied to X1 plus dot dot dot plus XL. And our plan will just be to Okay, so now this univariate polynomial Q um It's going to look like this.
- u281 Želimo, da je videti takole. Tukaj je L. To je slika polinoma ene spremenljivke, L minus 1, L deljeno z 2, 0. Aha, v redu. Ni važno. Pravzaprav bo plus L, minus L, 0, 1 ali morda 2, minus 2, nekaj takega.We want it to look like this. So here's L. This is a picture of a univariate polynomial L minus 1, L over 2, 0. Oh, okay. Never mind. Actually going to be plus L, minus L, 0, 1, or maybe 2, minus 2, something like this.
- u282 Em, parnost ... Ta čudni P za parnost, ki deluje za parnost, je natanko polinom, ki je morda plus 1 pri 0, potem minus 1 pri 2, potem plus 1 in minus 1, plus 1, minus 1.Um parity Okay, this funny parity P that works for parity, it's exactly a polynomial that's like maybe plus 1 at 0, and then minus 1 at 2, and then plus 1, and minus 1, plus 1, minus 1.
- u283 Videti je takole. To je P za parnost in je precej nora funkcija. V redu, to je P za parnost. Je neki polinom ene spremenljivke stopnje L, ki gre natanko skozi plus 1, potem minus 1, potem plus 1, potem minus 1, potem plus 1, potem minus 1.It looks like this. Like this is the the P for parity, and it's like a kind of crazy function. Like Okay, this is the P for parity. It's some univariate polynomial of degree L that exactly goes through like plus 1, then minus 1, then plus 1, then minus 1, then plus 1, then minus 1.
- u284 In vidite, da ko v P vstavite nekaj teh X I-jev, dobite polinom več spremenljivk, ki je natanko, saj veste, plus 1, minus 1, plus 1, minus 1, plus 1, minus 1. In je povsod popoln. No, povsod je parnost. Em, v redu, poskusili bomo narediti nekaj podobnega, le da bomo uporabili ta ... Uporabili bomo Q, ki mora imeti precej nižjo stopnjo, ampak upam, da je načrt očiten.And you see that like when you plug in some of the X I's into P, uh you get a multivariate polynomial that exactly is, you know, plus 1, minus 1, plus 1, minus 1, plus 1, minus 1. And it's perfect everywhere. Well, it is parity everywhere. Um Okay, and so we're going to try to do something similar, except we're going to use this uh We're going to use a Q um that has to have much less degree, but I hope the plan is obvious.
- u285 Te pluse ali minuse 1 bomo pravilno zadeli samo v tem okvirju tukaj s širino, saj veste, D je enako koren L log L. Torej želimo polinom Q, polinom ene spremenljivke, ki se ujema s tem plus 1, minus 1, plus 1, minus 1, plus 1, minus 1 samo v tem okvirju širine D.We're just going to get like these plus or minus 1's, right, on like this box here of width, you know, D equals root L log L. Okay, so we want a polynomial Q, a univariate polynomial that like matches this like plus 1, minus 1, plus 1, minus 1, plus 1, minus 1 just on this box of width D.
- u286 In spet, za poljubno množico D ali D plus 1 ali D minus 1 točk, kolikor jih pač želite, lahko natančno interpolirate s polinomom ene spremenljivke stopnje D. Tako bo tisti za Q natanko zadel vse te točke.And again, you know, for any set of D or D plus 1 or D minus 1 many points you want, you can interpolate it exactly with a univariate polynomial of degree D. And so it'll, you know, the one for Q will uh exactly hit all these points.
- u287 Ko ste zunaj tega okvirja, bo morda zavil v drugo smer kot P za parnost, ampak to nas ne zanima. Bistvo je, da bo Q res imel stopnjo samo koren log L. In s parnostjo se bo ujemal na vseh teh srednjih plasteh. Potem pa ima lahko zunaj tukaj povsem poljubne vrednosti.It'll go off in like maybe in a different direction from the P for parity once you're outside this box, but we don't care. The point is that Q will indeed have degree only root log L. And it'll match parity on all these like middle layers. Okay, then it might have like completely arbitrary values like outside here.
- u288 Ampak te nas ne zanimajo, ker predstavljajo samo delež ena deljeno s poli L kocke. Torej, odličen trenutek za povzetek. Nismo torej videli, kako to narediti.Uh but we don't care about them cuz they're only like one over poly L fraction of the cube. So, great time to sum up. Uh So, we didn't see how to do it.
- u289 Mislim, to je še vedno načrt. In ta načrt bomo uresničili s tem ključnim splošno znanim dejstvom. Em, [prhne] ali lahko o tem kaj povem v približno 1 minuti? Morda ne bom zares poskušal kaj povedati o tem v 1 minuti.I mean, this is still the plan. And this plan is going to be achieved by using this key folklore fact. Um [snorts] is there anything I can say about it in like 1 minute? Uh Maybe I will not really try to say anything about it in 1 minute.
- u290 No, v redu, povedal bom eno malenkost. Samo eno malenkost o tem, kakšni so ti AI-ji.Well, okay, I'll say one tiny thing. Just say one tiny thing about uh what do these AIs look like?
- u291 >> [prhne]>> [snorts]
- u292 >> Em, težko je sploh povedati, ampak čisto naslednja stvar, ki bi jo naredili, če bi imel več časa, je ...>> Uh it's hard to even say, but like the very next thing we would do if I had more time is
- u293 >> [prhne]>> [snorts]
- u294 >> Iz tega G bi sestavil nekaj zelo lepih matrik. Posebej bi sestavil eno lepo matriko, [prhne] Hadamardovo matriko.>> I would make some very nice matrices out of this G. In particular, like one nice matrix [snorts] I would make is a Hadamard matrix.
- u295 Em, indeksirano z 0 1 na L oziroma dva na L, krat diagonalna matrika z G krat Hadamardova matrika. To preprosto pomeni, da resničnostno tabelo G, ki ima spet poljubne realne vrednosti, postavim vzdolž diagonale.Uh like indexed by 0 1 to the L or 2 to the L times the diagonal matrix with G times the Hadamard matrix. And this just means that like I put the truth table of G which again is like arbitrary real values um along the diagonal.
- u296 In, em ...And uh
- u297 >> [prhne]>> [snorts]
- u298 >> Ta matrika bo zelo dober nadomestek za matriko, ki bi jo dobili, če bi na to diagonalo postavili pluse ali minuse parnosti. Nekako ima precej nižjo stopnjo.>> this matrix is going to be a very good surrogate for the matrix you would get if you put parities plus or minus entries down this diagonal. Somehow it's much lower degree.
- u299 >> [prhne]>> [snorts]
- u300 >> In če na to diagonalo postavite elemente plus ali minus ena iz parnosti ter z obeh strani pomnožite s to matriko Hadamardove Fourierove transformacije, dobite antidiagonalno matriko. Kar je pomembno.>> And if you put parities plus or minus one entries down this diagonal and multiply by this Hadamard Fourier transform matrix on both sides uh you get like the anti-diagonal matrix. Um which is important.
- u301 Navdušenci nad kvantnim računalništvom bodo to prepoznali kot H Z H je enako X v neki tenzorski obliki. Kakorkoli, nekako vzamete svojo A in jo faktorizirate tako, da se pojavi ta antidiagonalna matrika.Uh Quantum enthusiasts will like recognize this as like H Z H equals X in some tensor form. And anyway, like you sort of take your A factor it in such a way that this anti-diagonal matrix shows up.
- u302 Potem pa to antidiagonalno matriko zamenjate z nekimi približki, ki so videti takole, zaradi česar dobite bolj strukturirane matrike. Zmanjkalo mi je časa. Tega ne morem prav dobro razložiti. [smeh] To je nekaj sestavin. V redu.But then replace this anti-diagonal matrix by some approximations that look like this which causes you to get some more structured matrices. I'm out of time. I can't explain it very well. [laughter] Uh that's some of the ingredients. Okay.
- u303 [prhne] Končujemo torej zmedeno, ampak ali ima kdo kakšno vprašanje o dokazu ali rezultatih ali čemerkoli? Ja.[snorts] That's ending on a confusing note, but uh any questions about the proof or the results or anything? Yeah.
- u304 >> Em, boste dokaz dokončali naslednjo uro?>> Uh will you finish the proof next class?
- u305 >> Ali bom dokaz dokončal naslednjo uro? To je čudovito vprašanje. Ne vem. Pravzaprav naslednja ura odpade, kajne? Demokracija je. Ne, ja, dan demokracije. Šli bomo volit. Torej to morda poveča možnosti, ker bom imel en teden. Ne vem.>> Will I finish the proof next class? That's a wonderful question. I don't know. Actually, next class is off, right? It's democracy. No, yeah, democracy day. We're going to vote. So, maybe that increases the chances because I'll have a week. I don't know.
- u306 >> [smeh]>> [laughter]
- u307 >> NE VEM. ODVISNO JE od vas. Bi radi, da naslednjo uro dokončam dokaz, ali naj se vrnemo k ... V redu, vidim veliko prikimavanja. V redu, naslednjo uro bom dokončal dokaz. Zveni dobro.>> I DON'T KNOW. IT'S UP to you. Would you like me to finish the proof next class or go back to Okay, I'm getting a lot of nods. Okay, I'll finish proof next class. Sounds good.
V okolici
#3-SAT
Problem štetja izpolnjujočih prireditev logične formule v konjunktivni normalni obliki z največ tremi literali v vsaki klavzuli.
Tema1 virEn sam vir
Permanenta
Matrična funkcija, dobljena s seštevanjem produktov elementov, izbranih s permutacijami, brez predznakov permutacij, ki se uporabljajo pri determinanti.
Tema1 virEn sam vir
Polinomska aproksimacija logičnih funkcij
Predstavitev logične funkcije s polinomom ob določeni dopustni napaki.
Tema1 virEn sam vir
Vsota bločno diagonalne matrike in matrike nizkega ranga
Matrika, predstavljena kot vsota bločno diagonalne matrike in popravka nizkega ranga.
Tema1 virEn sam vir
Hipoteza o eksponentnem času
Hipoteza, da problema 3-SAT ni mogoče rešiti v času, ki je podeksponenten glede na število spremenljivk.
Tema1 virEn sam vir
Hermitova interpolacija
Polinomska interpolacija, ki predpisuje tako vrednosti funkcije kot vrednosti odvodov v izbranih točkah.
Tema1 virEn sam vir
Matrika nizkega ranga
Matrika, katere rang je majhen glede na njene dimenzije.
Tema1 virEn sam vir
Paritetna funkcija
Logična funkcija, ki pove, ali je število vhodnih bitov z vrednostjo ena sodo ali liho.
Tema1 virEn sam vir
To besedilo je napisano z umetno inteligenco.
Objavljeno v: Zdaj