Compressibility and probabilistic proofs alexander.shen@lirmm.fr, - - PowerPoint PPT Presentation

compressibility and probabilistic proofs
SMART_READER_LITE
LIVE PREVIEW

Compressibility and probabilistic proofs alexander.shen@lirmm.fr, - - PowerPoint PPT Presentation

Compressibility and probabilistic proofs alexander.shen@lirmm.fr, www.lirmm.fr/~ashen LIRMM CNRS & University of Montpellier CiE 2017 alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs Probabilistic


slide-1
SLIDE 1

Compressibility and probabilistic proofs

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen

LIRMM CNRS & University of Montpellier

CiE 2017

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-2
SLIDE 2

Probabilistic existence proofs

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-3
SLIDE 3

Probabilistic existence proofs

An object with some properties exists. . .

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-4
SLIDE 4

Probabilistic existence proofs

An object with some properties exists. . . because a random

  • bject has these properties (with positive probability)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-5
SLIDE 5

Probabilistic existence proofs

An object with some properties exists. . . because a random

  • bject has these properties (with positive probability)

A noncomputable binary sequence exists. . .

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-6
SLIDE 6

Probabilistic existence proofs

An object with some properties exists. . . because a random

  • bject has these properties (with positive probability)

A noncomputable binary sequence exists. . . . . . because the probability for a random sequence of fair coin tossings to be computable is 0

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-7
SLIDE 7

Probabilistic existence proofs

An object with some properties exists. . . because a random

  • bject has these properties (with positive probability)

A noncomputable binary sequence exists. . . . . . because the probability for a random sequence of fair coin tossings to be computable is 0 . . . because the probability of a random sequence to be computed by a given algorithm is 0 and we have countably many algorithms.

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-8
SLIDE 8

Probabilistic existence proofs

An object with some properties exists. . . because a random

  • bject has these properties (with positive probability)

A noncomputable binary sequence exists. . . . . . because the probability for a random sequence of fair coin tossings to be computable is 0 . . . because the probability of a random sequence to be computed by a given algorithm is 0 and we have countably many algorithms. cardinality argument in disguise, but we immediately get. . .

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-9
SLIDE 9

Probabilistic existence proofs

An object with some properties exists. . . because a random

  • bject has these properties (with positive probability)

A noncomputable binary sequence exists. . . . . . because the probability for a random sequence of fair coin tossings to be computable is 0 . . . because the probability of a random sequence to be computed by a given algorithm is 0 and we have countably many algorithms. cardinality argument in disguise, but we immediately get. . . Kleene, Post: there are non-comparable Turing degrees, i.e., two binary sequences that do not compute each other (being used as oracles)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-10
SLIDE 10

Probabilistic existence proofs

An object with some properties exists. . . because a random

  • bject has these properties (with positive probability)

A noncomputable binary sequence exists. . . . . . because the probability for a random sequence of fair coin tossings to be computable is 0 . . . because the probability of a random sequence to be computed by a given algorithm is 0 and we have countably many algorithms. cardinality argument in disguise, but we immediately get. . . Kleene, Post: there are non-comparable Turing degrees, i.e., two binary sequences that do not compute each other (being used as oracles) The probability of “β computes α” for random independent α and β is zero (fixed β + Fubini’s theorem), and vice versa

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-11
SLIDE 11

Finite probabilistic existence proof

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-12
SLIDE 12

Finite probabilistic existence proof

Boolean matrices n × n

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-13
SLIDE 13

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-14
SLIDE 14

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns monochromatic minor: all zeros/all ones

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-15
SLIDE 15

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns monochromatic minor: all zeros/all ones Theorem: for k = 3 log n and large n there exists a matrix without k × k monochromatic minors

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-16
SLIDE 16

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns monochromatic minor: all zeros/all ones Theorem: for k = 3 log n and large n there exists a matrix without k × k monochromatic minors Proof: for a random matrix the probability to have a large monochromatic minor is small (and therefore < 1)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-17
SLIDE 17

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns monochromatic minor: all zeros/all ones Theorem: for k = 3 log n and large n there exists a matrix without k × k monochromatic minors Proof: for a random matrix the probability to have a large monochromatic minor is small (and therefore < 1) the probability to have a k × k monochromatic minor at a given position: 2 × 2−k2

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-18
SLIDE 18

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns monochromatic minor: all zeros/all ones Theorem: for k = 3 log n and large n there exists a matrix without k × k monochromatic minors Proof: for a random matrix the probability to have a large monochromatic minor is small (and therefore < 1) the probability to have a k × k monochromatic minor at a given position: 2 × 2−k2 number of possible positions: ≤ nk × nk = 22k log n

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-19
SLIDE 19

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns monochromatic minor: all zeros/all ones Theorem: for k = 3 log n and large n there exists a matrix without k × k monochromatic minors Proof: for a random matrix the probability to have a large monochromatic minor is small (and therefore < 1) the probability to have a k × k monochromatic minor at a given position: 2 × 2−k2 number of possible positions: ≤ nk × nk = 22k log n 2k log n ≪ k2 if k ≫ 2 log n, so the union bound works

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-20
SLIDE 20

Finite probabilistic existence proof

Boolean matrices n × n k × k minor: fix arbitrary k rows and k columns monochromatic minor: all zeros/all ones Theorem: for k = 3 log n and large n there exists a matrix without k × k monochromatic minors Proof: for a random matrix the probability to have a large monochromatic minor is small (and therefore < 1) the probability to have a k × k monochromatic minor at a given position: 2 × 2−k2 number of possible positions: ≤ nk × nk = 22k log n 2k log n ≪ k2 if k ≫ 2 log n, so the union bound works Just counting (of course)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-21
SLIDE 21

Same proof using the compression language

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-22
SLIDE 22

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-23
SLIDE 23

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string most strings are incompressible (cannot be described by fewer bits)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-24
SLIDE 24

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string most strings are incompressible (cannot be described by fewer bits) if matrix with a k × k monochromatic minor for k ≫ 2 log n is compressible

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-25
SLIDE 25

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string most strings are incompressible (cannot be described by fewer bits) if matrix with a k × k monochromatic minor for k ≫ 2 log n is compressible why? it has a short description:

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-26
SLIDE 26

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string most strings are incompressible (cannot be described by fewer bits) if matrix with a k × k monochromatic minor for k ≫ 2 log n is compressible why? it has a short description: each of 2k rows/columns of the minor requires log n bits, 2k log n in total

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-27
SLIDE 27

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string most strings are incompressible (cannot be described by fewer bits) if matrix with a k × k monochromatic minor for k ≫ 2 log n is compressible why? it has a short description: each of 2k rows/columns of the minor requires log n bits, 2k log n in total

  • ne bit for the color of the minor

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-28
SLIDE 28

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string most strings are incompressible (cannot be described by fewer bits) if matrix with a k × k monochromatic minor for k ≫ 2 log n is compressible why? it has a short description: each of 2k rows/columns of the minor requires log n bits, 2k log n in total

  • ne bit for the color of the minor

the rest of the matrix (n2 − k2 bits)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-29
SLIDE 29

Same proof using the compression language

n × n matrix can be encoded as a n2-bit string most strings are incompressible (cannot be described by fewer bits) if matrix with a k × k monochromatic minor for k ≫ 2 log n is compressible why? it has a short description: each of 2k rows/columns of the minor requires log n bits, 2k log n in total

  • ne bit for the color of the minor

the rest of the matrix (n2 − k2 bits) replacing k2 by 2k log n + 1: compression if k ≫ 2 log n

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-30
SLIDE 30

So what?

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-31
SLIDE 31

So what?

may be the compression language is more intuitive

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-32
SLIDE 32

So what?

may be the compression language is more intuitive but not very impressive. . .

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-33
SLIDE 33

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-34
SLIDE 34

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-35
SLIDE 35

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound algorithmic version due to Moses-Tardos

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-36
SLIDE 36

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound algorithmic version due to Moses-Tardos do not need to know what is LL and MT algorithm

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-37
SLIDE 37

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound algorithmic version due to Moses-Tardos do not need to know what is LL and MT algorithm scheme: we try to most natural randomized algorithm

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-38
SLIDE 38

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound algorithmic version due to Moses-Tardos do not need to know what is LL and MT algorithm scheme: we try to most natural randomized algorithm it succeeds with high probability. . .

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-39
SLIDE 39

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound algorithmic version due to Moses-Tardos do not need to know what is LL and MT algorithm scheme: we try to most natural randomized algorithm it succeeds with high probability. . . because if it fails, the random bits used are compressible

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-40
SLIDE 40

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound algorithmic version due to Moses-Tardos do not need to know what is LL and MT algorithm scheme: we try to most natural randomized algorithm it succeeds with high probability. . . because if it fails, the random bits used are compressible A: forbidden factors (Ochem, Gon¸ calves)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-41
SLIDE 41

So what?

may be the compression language is more intuitive but not very impressive. . . more interesting examples Lovasz local lemma instead of the union bound algorithmic version due to Moses-Tardos do not need to know what is LL and MT algorithm scheme: we try to most natural randomized algorithm it succeeds with high probability. . . because if it fails, the random bits used are compressible A: forbidden factors (Ochem, Gon¸ calves) B: CNF with bounded neighborhood (Moser, Fortnow)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-42
SLIDE 42

Forbidden factors

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-43
SLIDE 43

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-44
SLIDE 44

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”) is there an infinite bit sequence that does not have any of Fi as a substring?

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-45
SLIDE 45

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”) is there an infinite bit sequence that does not have any of Fi as a substring? infinite ⇔ arbitrarily long

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-46
SLIDE 46

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”) is there an infinite bit sequence that does not have any of Fi as a substring? infinite ⇔ arbitrarily long the answer depends on the list: 0, 11 does not exist;

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-47
SLIDE 47

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”) is there an infinite bit sequence that does not have any of Fi as a substring? infinite ⇔ arbitrarily long the answer depends on the list: 0, 11 does not exist; 0, 00 does exist

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-48
SLIDE 48

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”) is there an infinite bit sequence that does not have any of Fi as a substring? infinite ⇔ arbitrarily long the answer depends on the list: 0, 11 does not exist; 0, 00 does exist for a fixed list we get a regular expression / finite automaton

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-49
SLIDE 49

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”) is there an infinite bit sequence that does not have any of Fi as a substring? infinite ⇔ arbitrarily long the answer depends on the list: 0, 11 does not exist; 0, 00 does exist for a fixed list we get a regular expression / finite automaton quantitative results: “if there are not too many forbidden strings of each length, then there are long sequences without forbidden strings”

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-50
SLIDE 50

Forbidden factors

F1, . . . , Fk: binary strings (“forbidden strings”) is there an infinite bit sequence that does not have any of Fi as a substring? infinite ⇔ arbitrarily long the answer depends on the list: 0, 11 does not exist; 0, 00 does exist for a fixed list we get a regular expression / finite automaton quantitative results: “if there are not too many forbidden strings of each length, then there are long sequences without forbidden strings” Let ai be the number of forbidden strings of length i. If

  • aiti < mt − 1 for some t > 0

then there exist arbitrarily long strings without forbidden

  • factors. (For the case of m letters)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-51
SLIDE 51

Tetris algorithm

Forbidden strings: 01, 110 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-52
SLIDE 52

Tetris algorithm

Forbidden strings: 01, 110 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-53
SLIDE 53

Tetris algorithm

Forbidden strings: 01, 110 00 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-54
SLIDE 54

Tetris algorithm

Forbidden strings: 01, 110

1

001 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-55
SLIDE 55

Tetris algorithm

Forbidden strings: 01, 110

1

001 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-56
SLIDE 56

Tetris algorithm

Forbidden strings: 01, 110 001 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-57
SLIDE 57

Tetris algorithm

Forbidden strings: 01, 110

1

0011 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-58
SLIDE 58

Tetris algorithm

Forbidden strings: 01, 110

1

0011 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-59
SLIDE 59

Tetris algorithm

Forbidden strings: 01, 110 0011 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-60
SLIDE 60

Tetris algorithm

Forbidden strings: 01, 110

1

00111 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-61
SLIDE 61

Tetris algorithm

Forbidden strings: 01, 110

1 1

001111 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-62
SLIDE 62

Tetris algorithm

Forbidden strings: 01, 110

1 1

0011110 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-63
SLIDE 63

Tetris algorithm

Forbidden strings: 01, 110

1 1

0011110 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-64
SLIDE 64

Tetris algorithm

Forbidden strings: 01, 110 0011110 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-65
SLIDE 65

Tetris algorithm

Forbidden strings: 01, 110

1

00111101 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-66
SLIDE 66

Tetris algorithm

Forbidden strings: 01, 110

1

001111010 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-67
SLIDE 67

Tetris algorithm

Forbidden strings: 01, 110

1

0011110100 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-68
SLIDE 68

Tetris algorithm

Forbidden strings: 01, 110

1

00111101000 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-69
SLIDE 69

Tetris algorithm

Forbidden strings: 01, 110

1 1

001111010001 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-70
SLIDE 70

Tetris algorithm

Forbidden strings: 01, 110

1 1

001111010001 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-71
SLIDE 71

Tetris algorithm

Forbidden strings: 01, 110

1

001111010001 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-72
SLIDE 72

Tetris algorithm

Forbidden strings: 01, 110

1

0011110100010 Random bits are added one by one; if a forbidden string appears (at the end), it vanishes, and the process continues

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-73
SLIDE 73

Will it grow indefinitely?

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-74
SLIDE 74

Will it grow indefinitely?

alphabet size m

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-75
SLIDE 75

Will it grow indefinitely?

alphabet size m an is the number of forbidden strings of length n ≥ 2

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-76
SLIDE 76

Will it grow indefinitely?

alphabet size m an is the number of forbidden strings of length n ≥ 2 assume that

n antn < mt − 1 for some t > 0

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-77
SLIDE 77

Will it grow indefinitely?

alphabet size m an is the number of forbidden strings of length n ≥ 2 assume that

n antn < mt − 1 for some t > 0

Claim: if the string remains short forever, then the sequence

  • f random bits is compressible

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-78
SLIDE 78

Will it grow indefinitely?

alphabet size m an is the number of forbidden strings of length n ≥ 2 assume that

n antn < mt − 1 for some t > 0

Claim: if the string remains short forever, then the sequence

  • f random bits is compressible

log file: sequence of signs like +, +01, +110 for adding the new bit (not indicated) without or with cancelled string

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-79
SLIDE 79

Will it grow indefinitely?

alphabet size m an is the number of forbidden strings of length n ≥ 2 assume that

n antn < mt − 1 for some t > 0

Claim: if the string remains short forever, then the sequence

  • f random bits is compressible

log file: sequence of signs like +, +01, +110 for adding the new bit (not indicated) without or with cancelled string going backwards: + means deletion of the last bit, +u means adding u and then deleting the last bit

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-80
SLIDE 80

Will it grow indefinitely?

alphabet size m an is the number of forbidden strings of length n ≥ 2 assume that

n antn < mt − 1 for some t > 0

Claim: if the string remains short forever, then the sequence

  • f random bits is compressible

log file: sequence of signs like +, +01, +110 for adding the new bit (not indicated) without or with cancelled string going backwards: + means deletion of the last bit, +u means adding u and then deleting the last bit current sequence + log file → random bits used

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-81
SLIDE 81

Will it grow indefinitely?

alphabet size m an is the number of forbidden strings of length n ≥ 2 assume that

n antn < mt − 1 for some t > 0

Claim: if the string remains short forever, then the sequence

  • f random bits is compressible

log file: sequence of signs like +, +01, +110 for adding the new bit (not indicated) without or with cancelled string going backwards: + means deletion of the last bit, +u means adding u and then deleting the last bit current sequence + log file → random bits used few forbidden strings ⇒ few symbols in log file ⇒ efficient encoding

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-82
SLIDE 82

More details

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-83
SLIDE 83

More details

current string + log file → sequence of random bits

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-84
SLIDE 84

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-85
SLIDE 85

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-86
SLIDE 86

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits so we need to encode the log file efficiently (< 1 bit/symbol)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-87
SLIDE 87

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits so we need to encode the log file efficiently (< 1 bit/symbol) large alphabet +x but most symbols are +

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-88
SLIDE 88

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits so we need to encode the log file efficiently (< 1 bit/symbol) large alphabet +x but most symbols are + arithmetic coding: use less that 1 bit for +

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-89
SLIDE 89

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits so we need to encode the log file efficiently (< 1 bit/symbol) large alphabet +x but most symbols are + arithmetic coding: use less that 1 bit for + the savings due to +’s are used for encoding +x letters

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-90
SLIDE 90

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits so we need to encode the log file efficiently (< 1 bit/symbol) large alphabet +x but most symbols are + arithmetic coding: use less that 1 bit for + the savings due to +’s are used for encoding +x letters amortized analysis: + increases the length by 1 and +x decreases the length by |x| − 1

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-91
SLIDE 91

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits so we need to encode the log file efficiently (< 1 bit/symbol) large alphabet +x but most symbols are + arithmetic coding: use less that 1 bit for + the savings due to +’s are used for encoding +x letters amortized analysis: + increases the length by 1 and +x decreases the length by |x| − 1 so there couldn’t be many +x unless there are many +

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-92
SLIDE 92

More details

current string + log file → sequence of random bits current string is O(1) if it doesn’t grow indefinitely the length of log file is the number of random bits so we need to encode the log file efficiently (< 1 bit/symbol) large alphabet +x but most symbols are + arithmetic coding: use less that 1 bit for + the savings due to +’s are used for encoding +x letters amortized analysis: + increases the length by 1 and +x decreases the length by |x| − 1 so there couldn’t be many +x unless there are many + role of t: parameter for amortized analysis of the encoding efficiency

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-93
SLIDE 93

Technical details

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-94
SLIDE 94

Technical details

arithmetic coding: each symbol z has some weight pz > 0

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-95
SLIDE 95

Technical details

arithmetic coding: each symbol z has some weight pz > 0

  • z pz ≤ 1

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-96
SLIDE 96

Technical details

arithmetic coding: each symbol z has some weight pz > 0

  • z pz ≤ 1

encoding z by log(1/pz) bits

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-97
SLIDE 97

Technical details

arithmetic coding: each symbol z has some weight pz > 0

  • z pz ≤ 1

encoding z by log(1/pz) bits allocate weight q0 for + and total weight qn for all +u where u are forbidden strings of length n.

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-98
SLIDE 98

Technical details

arithmetic coding: each symbol z has some weight pz > 0

  • z pz ≤ 1

encoding z by log(1/pz) bits allocate weight q0 for + and total weight qn for all +u where u are forbidden strings of length n. code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-99
SLIDE 99

Technical details

arithmetic coding: each symbol z has some weight pz > 0

  • z pz ≤ 1

encoding z by log(1/pz) bits allocate weight q0 for + and total weight qn for all +u where u are forbidden strings of length n. code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n each log symbol corresponds to one random symbol, so we want to encode log symbols with less than log m bits

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-100
SLIDE 100

Technical details

arithmetic coding: each symbol z has some weight pz > 0

  • z pz ≤ 1

encoding z by log(1/pz) bits allocate weight q0 for + and total weight qn for all +u where u are forbidden strings of length n. code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n each log symbol corresponds to one random symbol, so we want to encode log symbols with less than log m bits + increases the length by 1, and +u decreases the length by n − 1 for |u| = n

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-101
SLIDE 101

Technical details

arithmetic coding: each symbol z has some weight pz > 0

  • z pz ≤ 1

encoding z by log(1/pz) bits allocate weight q0 for + and total weight qn for all +u where u are forbidden strings of length n. code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n each log symbol corresponds to one random symbol, so we want to encode log symbols with less than log m bits + increases the length by 1, and +u decreases the length by n − 1 for |u| = n amortized analysis: when increasing length (+), reserve δ; when decreasing length n − 1, use δ(n − 1) from reserves.

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-102
SLIDE 102

Technical details-2

code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n each log symbol corresponds to one random symbol, so we want to encode log symbols with less than log m bits + increases the length by 1, and +u decreases the length by n − 1 for |u| = n amortized analysis: when increasing length (+), reserve δ; when decreasing length n − 1, use δ(n − 1) from reserves.

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-103
SLIDE 103

Technical details-2

code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n each log symbol corresponds to one random symbol, so we want to encode log symbols with less than log m bits + increases the length by 1, and +u decreases the length by n − 1 for |u| = n amortized analysis: when increasing length (+), reserve δ; when decreasing length n − 1, use δ(n − 1) from reserves. − log q0 ≤ log m − δ

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-104
SLIDE 104

Technical details-2

code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n each log symbol corresponds to one random symbol, so we want to encode log symbols with less than log m bits + increases the length by 1, and +u decreases the length by n − 1 for |u| = n amortized analysis: when increasing length (+), reserve δ; when decreasing length n − 1, use δ(n − 1) from reserves. − log q0 ≤ log m − δ − log qn + log an ≤ log m + (n − 1)δ

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-105
SLIDE 105

Technical details-2

code lengths: − log q0, for “+” − log qn + log an, for each “+u” with |u| = n each log symbol corresponds to one random symbol, so we want to encode log symbols with less than log m bits + increases the length by 1, and +u decreases the length by n − 1 for |u| = n amortized analysis: when increasing length (+), reserve δ; when decreasing length n − 1, use δ(n − 1) from reserves. − log q0 ≤ log m − δ − log qn + log an ≤ log m + (n − 1)δ

  • n qn < 1

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-106
SLIDE 106

Technical details-3

− log q0 ≤ log m − δ − log qn + log an ≤ log m + (n − 1)δ

  • n qn < 1

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-107
SLIDE 107

Technical details-3

− log q0 = log m − δ − log qn + log an = log m + (n − 1)δ

  • n qn < 1

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-108
SLIDE 108

Technical details-3

− log q0 = log m − δ; q0 = (1/m)2δ − log qn + log an = log m + (n − 1)δ; qn = (1/m)an2δ2−nδ

  • n qn < 1

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-109
SLIDE 109

Technical details-3

− log q0 = log m − δ; q0 = (1/m)2δ − log qn + log an = log m + (n − 1)δ; qn = (1/m)an2δ2−nδ

  • n qn < 1

(1/m)2δ + (1/m)2δ

n an(2−δ)n < 1

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-110
SLIDE 110

Technical details-3

− log q0 = log m − δ; q0 = (1/m)2δ − log qn + log an = log m + (n − 1)δ; qn = (1/m)an2δ2−nδ

  • n qn < 1

(1/m)2δ + (1/m)2δ

n an(2−δ)n < 1

1 +

n an(2−δ)n < m(2−δ)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-111
SLIDE 111

Technical details-3

− log q0 = log m − δ; q0 = (1/m)2δ − log qn + log an = log m + (n − 1)δ; qn = (1/m)an2δ2−nδ

  • n qn < 1

(1/m)2δ + (1/m)2δ

n an(2−δ)n < 1

1 +

n an(2−δ)n < m(2−δ)

let t = 2−δ

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-112
SLIDE 112

Technical details-3

− log q0 = log m − δ; q0 = (1/m)2δ − log qn + log an = log m + (n − 1)δ; qn = (1/m)an2δ2−nδ

  • n qn < 1

(1/m)2δ + (1/m)2δ

n an(2−δ)n < 1

1 +

n an(2−δ)n < m(2−δ)

let t = 2−δ

  • n antn < mt − 1, as stated

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-113
SLIDE 113

CNF with bounded neighborhood

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-114
SLIDE 114

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-115
SLIDE 115

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-116
SLIDE 116

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal literal: propositional variable or its negation

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-117
SLIDE 117

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal literal: propositional variable or its negation (p ∨ q) ∧ (p ∨ ¬q) ∧ (¬p ∨ q) ∧ (¬p ∨ ¬q)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-118
SLIDE 118

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal literal: propositional variable or its negation (p ∨ q) ∧ (p ∨ ¬q) ∧ (¬p ∨ q) ∧ (¬p ∨ ¬q) each clause prohibits some combination of values

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-119
SLIDE 119

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal literal: propositional variable or its negation (p ∨ q) ∧ (p ∨ ¬q) ∧ (¬p ∨ q) ∧ (¬p ∨ ¬q) each clause prohibits some combination of values here all four combinations are prohibited, unsatisfiable

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-120
SLIDE 120

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal literal: propositional variable or its negation (p ∨ q) ∧ (p ∨ ¬q) ∧ (¬p ∨ q) ∧ (¬p ∨ ¬q) each clause prohibits some combination of values here all four combinations are prohibited, unsatisfiable Assume all clauses are with n literals, thus prohibiting one combination for some n variables.

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-121
SLIDE 121

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal literal: propositional variable or its negation (p ∨ q) ∧ (p ∨ ¬q) ∧ (¬p ∨ q) ∧ (¬p ∨ ¬q) each clause prohibits some combination of values here all four combinations are prohibited, unsatisfiable Assume all clauses are with n literals, thus prohibiting one combination for some n variables. To make the CNF unsatisfiable, we need about 2n of them and they should have more or less the same variables:

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-122
SLIDE 122

CNF with bounded neighborhood

CNF: clause ∧ clause ∧ . . . ∧ clause clause: literal ∨ literal ∨ . . . ∨ literal literal: propositional variable or its negation (p ∨ q) ∧ (p ∨ ¬q) ∧ (¬p ∨ q) ∧ (¬p ∨ ¬q) each clause prohibits some combination of values here all four combinations are prohibited, unsatisfiable Assume all clauses are with n literals, thus prohibiting one combination for some n variables. To make the CNF unsatisfiable, we need about 2n of them and they should have more or less the same variables: Claim: If each clause has n literals and has at most 2n−3 neighbors (=clauses that have common variable), then CNF is satisfiable.

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-123
SLIDE 123

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true }

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-124
SLIDE 124

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } this is enough (fixing clauses one by one)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-125
SLIDE 125

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } this is enough (fixing clauses one by one) conditional correctness

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-126
SLIDE 126

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } this is enough (fixing clauses one by one) conditional correctness termination?

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-127
SLIDE 127

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } this is enough (fixing clauses one by one) conditional correctness termination? Claim: if no termination after a long time, the sequence of random bits used for resampling is compressible

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-128
SLIDE 128

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true }

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-129
SLIDE 129

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } long (unfinished) execution of Fix(C)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-130
SLIDE 130

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } long (unfinished) execution of Fix(C) log file: list of clauses for all calls Fix(C ′)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-131
SLIDE 131

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } long (unfinished) execution of Fix(C) log file: list of clauses for all calls Fix(C ′)

  • nly false clauses are fixed

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-132
SLIDE 132

Fixing clauses

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } long (unfinished) execution of Fix(C) log file: list of clauses for all calls Fix(C ′)

  • nly false clauses are fixed

knowing this list, and the current values of variable we can go backwards and reconstruct the values of variables and bits used for resampling — and the log file is the compressed encoding of the bits used for the resampling

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-133
SLIDE 133

Tree traversal

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true }

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-134
SLIDE 134

Tree traversal

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } main observation: C ′ is a neighbor of C

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-135
SLIDE 135

Tree traversal

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } main observation: C ′ is a neighbor of C in the tree of recursive calls sons are neighbors

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-136
SLIDE 136

Tree traversal

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } main observation: C ′ is a neighbor of C in the tree of recursive calls sons are neighbors we specify a neighbor using n − 3 bits instead of n needed to specify resampling bits: compression

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-137
SLIDE 137

Tree traversal

{ C is false } Fix(C : clause): Resample(C) for all C ′ that are neighbors of C (including C): if C ′ is false then Fix(C ′) { C is true; all clauses that were true remain true } main observation: C ′ is a neighbor of C in the tree of recursive calls sons are neighbors we specify a neighbor using n − 3 bits instead of n needed to specify resampling bits: compression techically incorrect, since we also go down the tree (return from recursive calls) - we need to reserve two more bits ((n − 3) + 2 < n)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-138
SLIDE 138

Tree traversal: counting bits

C1 C2 C3 C4 C5 C6 C7 C8 C9

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-139
SLIDE 139

Tree traversal: counting bits

C1 C2 C3 C4 C5 C6 C7 C8 C9

n − 3 bits: neighbor number

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-140
SLIDE 140

Tree traversal: counting bits

C1 C2 C3 C4 C5 C6 C7 C8 C9

n − 3 bits: neighbor number plus 1 direction bit “up” (when going up)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-141
SLIDE 141

Tree traversal: counting bits

C1 C2 C3 C4 C5 C6 C7 C8 C9

n − 3 bits: neighbor number plus 1 direction bit “up” (when going up) 1 direction bit (when going down)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-142
SLIDE 142

Tree traversal: counting bits

C1 C2 C3 C4 C5 C6 C7 C8 C9

n − 3 bits: neighbor number plus 1 direction bit “up” (when going up) 1 direction bit (when going down) (n − 3) + 1 + 1 per one move up (n sampling bits): (n − 1) instead of n

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-143
SLIDE 143

History and references

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-144
SLIDE 144

History and references

Probabilistic/averaging arguments — Littlewood (Mathematical miscellany?)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-145
SLIDE 145

History and references

Probabilistic/averaging arguments — Littlewood (Mathematical miscellany?) Lovasz local lemma (1975)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-146
SLIDE 146

History and references

Probabilistic/averaging arguments — Littlewood (Mathematical miscellany?) Lovasz local lemma (1975) Moser (2008)–Tardos (2009)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-147
SLIDE 147

History and references

Probabilistic/averaging arguments — Littlewood (Mathematical miscellany?) Lovasz local lemma (1975) Moser (2008)–Tardos (2009) Miller (potential, ≤ 2011)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-148
SLIDE 148

History and references

Probabilistic/averaging arguments — Littlewood (Mathematical miscellany?) Lovasz local lemma (1975) Moser (2008)–Tardos (2009) Miller (potential, ≤ 2011) Golod–Shafarevich (1964)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-149
SLIDE 149

History and references

Probabilistic/averaging arguments — Littlewood (Mathematical miscellany?) Lovasz local lemma (1975) Moser (2008)–Tardos (2009) Miller (potential, ≤ 2011) Golod–Shafarevich (1964) Ochem, Gon¸ calves (2014)

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs

slide-150
SLIDE 150

History and references

Probabilistic/averaging arguments — Littlewood (Mathematical miscellany?) Lovasz local lemma (1975) Moser (2008)–Tardos (2009) Miller (potential, ≤ 2011) Golod–Shafarevich (1964) Ochem, Gon¸ calves (2014)

Thanks for the attention!

alexander.shen@lirmm.fr, www.lirmm.fr/~ashen Compressibility and probabilistic proofs