Bounded-depth circuits: Separating wires from gates Michal Koucký Joint work with: Pavel Pudlák and Denis Thérien.

Slides:



Advertisements
Similar presentations
TWO STEP EQUATIONS 1. SOLVE FOR X 2. DO THE ADDITION STEP FIRST
Advertisements

There is a pattern for factoring trinomials of this form, when c
By D. Fisher Geometric Transformations. Reflection, Rotation, or Translation 1.
Factors, Primes & Composite Numbers
Analysis of Algorithms
Jeopardy Q 1 Q 6 Q 11 Q 16 Q 21 Q 2 Q 7 Q 12 Q 17 Q 22 Q 3 Q 8 Q 13
Jeopardy Q 1 Q 6 Q 11 Q 16 Q 21 Q 2 Q 7 Q 12 Q 17 Q 22 Q 3 Q 8 Q 13
Chapter R: Reference: Basic Algebraic Concepts
0 - 0.
ALGEBRAIC EXPRESSIONS
DIVIDING INTEGERS 1. IF THE SIGNS ARE THE SAME THE ANSWER IS POSITIVE 2. IF THE SIGNS ARE DIFFERENT THE ANSWER IS NEGATIVE.
MULTIPLYING MONOMIALS TIMES POLYNOMIALS (DISTRIBUTIVE PROPERTY)
ADDING INTEGERS 1. POS. + POS. = POS. 2. NEG. + NEG. = NEG. 3. POS. + NEG. OR NEG. + POS. SUBTRACT TAKE SIGN OF BIGGER ABSOLUTE VALUE.
MULTIPLICATION EQUATIONS 1. SOLVE FOR X 3. WHAT EVER YOU DO TO ONE SIDE YOU HAVE TO DO TO THE OTHER 2. DIVIDE BY THE NUMBER IN FRONT OF THE VARIABLE.
SUBTRACTING INTEGERS 1. CHANGE THE SUBTRACTION SIGN TO ADDITION
MULT. INTEGERS 1. IF THE SIGNS ARE THE SAME THE ANSWER IS POSITIVE 2. IF THE SIGNS ARE DIFFERENT THE ANSWER IS NEGATIVE.
FACTORING Think Distributive property backwards Work down, Show all steps ax + ay = a(x + y)
Addition Facts
ALGEBRAIC EXPRESSIONS
Reductions Complexity ©D.Moshkovitz.
Solve Multi-step Equations
Mathematics for Economics Beatrice Venturi 1 Economics Faculty CONTINUOUS TIME: LINEAR DIFFERENTIAL EQUATIONS Economic Applications LESSON 2 prof. Beatrice.
Long cycles, short cycles, min-degree subgraphs, and feedback arc sets in Eulerian digraphs Raphael Yuster joint work with Asaf Shapira Eilat 2012.
Chapter 9 -- Simplification of Sequential Circuits.
Gates and Circuits Nell Dale & John Lewis (adaptation by Erin Chambers and Michael Goldwasser)
Digital Logic Design Gate-Level Minimization
Introduction to Logic Gates
Exclusive-OR and Exclusive-NOR Gates
1 Computer Programming Boolean Logic Copyright © Texas Education Agency, 2013.
Charles Kime & Thomas Kaminski © 2008 Pearson Education, Inc. (Hyperlinks are active in View Show mode) Chapter 2 – Combinational Logic Circuits Part 3.
Warm-up Divide the following using Long Division:
O X Click on Number next to person for a question.
© S Haughton more than 3?
Solving Equations How to Solve Them
1 Directed Depth First Search Adjacency Lists A: F G B: A H C: A D D: C F E: C D G F: E: G: : H: B: I: H: F A B C G D E H I.
Twenty Questions Subject: Twenty Questions
1 SIP 250, AB We’ll use reductions that, by default, would be of this type, which is called: Polynomial-time mapping reduction Polynomial-time.
Linking Verb? Action Verb or. Question 1 Define the term: action verb.
Past Tense Probe. Past Tense Probe Past Tense Probe – Practice 1.
This, that, these, those Number your paper from 1-10.
Chapter 11 Polynomials.
Properties of Exponents
Event 4: Mental Math 7th/8th grade Math Meet ‘11.
Addition 1’s to 20.
25 seconds left…...
Test B, 100 Subtraction Facts
11 = This is the fact family. You say: 8+3=11 and 3+8=11
Week 1.
We will resume in: 25 Minutes.
O X Click on Number next to person for a question.
Chapter 11 Limitations of Algorithm Power Copyright © 2007 Pearson Addison-Wesley. All rights reserved.
Bart Jansen 1.  Problem definition  Instance: Connected graph G, positive integer k  Question: Is there a spanning tree for G with at least k leaves?
Circuit and Communication Complexity. Karchmer – Wigderson Games Given The communication game G f : Alice getss.t. f(x)=1 Bob getss.t. f(y)=0 Goal: Find.
Equivalence Relations
Discrete Structures Chapter 6: Set Theory
Succinct Data Structures for Permutations, Functions and Suffix Arrays
12-Apr-15 Analysis of Algorithms. 2 Time and space To analyze an algorithm means: developing a formula for predicting how fast an algorithm is, based.
Amplifying lower bounds by means of self- reducibility Eric Allender Michal Koucký Rutgers University Academy of Sciences Czech Republic Czech Republic.
Finite monoids, regular languages, circuit complexity and logic Pascal Tesson Laval University, Quebec City, Canada.
CSCI 4325 / 6339 Theory of Computation Zhixiang Chen.
Circuit Complexity of Regular Languages Michal Koucký (Institute of Mathemaics, AS ČR, Praha)
Circuit Lower Bounds via Ehrenfeucht- Fraïssé Games Michal Koucký Joint work with: Clemens Lautemann, Sebastian Poloczek, Denis Thérien.
Combination of logic gates  Logic gates can be combined to produce more complex functions.  They can also be combined to substitute one type of gate.
Linear Algebra. Circuits The circuits in computers and other input devices have inputs, each of which is either a 0 or 1, the output is also 0s and 1s.
NP-Completeness (2) NP-Completeness Graphs 4/13/2018 5:22 AM x x x x x
A new characterization of ACC0 and probabilistic CC0
Circuit Lower Bounds A combinatorial approach to P vs NP
The story of superconcentrators The missing link
NP-Completeness (2) NP-Completeness Graphs 7/9/2019 6:12 AM x x x x x
Presentation transcript:

Bounded-depth circuits: Separating wires from gates Michal Koucký Joint work with: Pavel Pudlák and Denis Thérien

2 Boolean circuits: x 1 x 2 x 4 x 7 x 1 x 2 x 4 x 7 non-uniform model of time bounded computation. non-uniform model of time bounded computation. Fundamental question: How large circuits does one need to compute specific Boolean functions, e.g., SAT?

3 Size of circuits = number of gates? = number of wires? Different applications use different measures counting circuits counting circuits first order logic first order logic oracle gates oracle gates Question: Arent these two measures indeed the same?

4 Answer: By counting, these two measures are different: circuits with f ( n ) gates can compute functions which no circuit with O( f ( n )) wires can compute. circuits with f ( n ) gates can compute functions which no circuit with O( f ( n )) wires can compute. Question: Is there an interesting (simple and explicit) function which demonstrates the difference?

5 Our results: Explicit function computable by AC 0 circuits with O(n ) gates but not O( n ) wires. Explicit function computable by AC 0 circuits with O(n ) gates but not O( n ) wires. Same for AC 0 [ q ] and ACC 0 circuits. Same for AC 0 [ q ] and ACC 0 circuits. (ac* bc*)* (ac* bc*)* AC 0 circuits … constant-depth circuits consisting of polynomially many,, gates. AC 0 [ q ] circuits … contain in addition MOD-q gates. ACC 0 = q AC 0 [ q ].

6 Precise characterization of regular languages computable by AC 0 circuits with O( n ) wires. Precise characterization of regular languages computable by AC 0 circuits with O( n ) wires. … unambiguous languages ( DA ) Same for ACC 0. Same for ACC 0. … languages with syntactic monoids in DO Ab. New way to analyze communication in circuits. New way to analyze communication in circuits.

7 Related results: [RW91] … poly-log threshold in AC 0 with O(n) gates. [CJ96] … Ajtais AC 0 approximate counting is not in AC 0 with O(n) gates. [CFL83] … all regular languages that are in AC 0 require only O(n g O(d ) (n )) wires. g 1 ( n ) = log n g d+1 ( n ) = ( g d ( n ) )* g 1 ( n ) = log n g d+1 ( n ) = ( g d ( n ) )*

8 Communication in circuits

9 k, X, Y k, X, Y |X|=|Y|=k k disj. paths 0< ε 1 const. |In|=|Out|=n d Super-concentrators XY k k Out In Thm [DDPW, P]: Any super-concentrator of depth d contains at least (n g d (n )) wires. True even if there are many paths between X,Ys of special form only.

10 x 1 x 2 x 3 … d y 1 y 2 y 3 … Circuit lower bounds 2 k possible outputs 2 k possible inputs AC 0 [ ] circuit for PREFIX-PARITY. PREFIX-PARITY … i-th bit of output is parity of the first i bits of the input. k k

11 Any AC 0 [ ] circuit of depth d for PREFIX-PARITY contains (n g d (n )) wires. [CFL83] Any AC 0 circuit for integer ADDITION contains (n g d (n )) wires. x 1 x 2 x 3 … d y 1 y 2 y 3 … Circuit lower bounds 2 k possible outputs 2 k possible inputs k k

12 f (X,Y ) of large communication complexity every cut S that separates information coming from X and Y is large. every cut S that separates information coming from X and Y is large. X Y Boolean function S

13 (X,Y )-cut … set S of gates such that after removing S from circuit C there is no (X,Y )- path. (X,Y )-cut … set S of gates such that after removing S from circuit C there is no (X,Y )- path. f (X,Y ) has communication complexity k f (X,Y ) has communication complexity k every (X,Y )-cut in C is of size ( k ) every (X,Y )-cut in C is of size ( k ) ( k ) vertex disjoint (X,Y )-paths in C. ( k ) vertex disjoint (X,Y )-paths in C.

14 Circuit C n computing (ac* bc*)* x 1 x 2 x 3 … x 1 x 2 x 3 … a cc … ? … ? … cc a cc … ? … ? … c a c … ? … y 1 y 2 … y 1 y 2 … k equally spaced as k equally spaced as X = {x 1, x 2, x 3 …, x k } Y = {y 1, y 2, y 3 …, y k } C n has to compute bit-wise complement of x 1 x 2 x 3 …x k and y 1 y 2 y 3 …y k. C n has to compute bit-wise complement of x 1 x 2 x 3 …x k and y 1 y 2 y 3 …y k. ( k ) vertex disjoint (X,Y )-paths in C n. ( k ) vertex disjoint (X,Y )-paths in C n.

15 Open problems Is integer ADDITION in AC 0 with linear number of gates? Is integer ADDITION in AC 0 with linear number of gates? Is (ac* bc*)* in AC 0 with linear number of gates? Is (ac* bc*)* in AC 0 with linear number of gates?

16

17 X Turning C into a super-concentrator Y Y Y 2d+1 copies of C 2d+1 copies of C

18 DA satisfies: (x y z ) y (x y z ) = (x y z ) (x y z ) y (x y z ) = (x y z ) … turtle languages … disj. unions of unambiguous 1 *a 1 2 *a 2 … k * DO satisfies: (x y) ( y x ) (x y) = (x y) (x y) ( y x ) (x y) = (x y) DO Ab DO Ab … super-turtle languages … disj. unions of unambiguous L 1 a 1 L 2 a 2 … L k … exponent of M (how long one has to multiply x to get idempotent m 2 =m). … exponent of M (how long one has to multiply x to get idempotent m 2 =m).

19 Function in AC 0 computable with O(n ) gates but not with O(n ) wires: Function in AC 0 computable with O(n ) gates but not with O(n ) wires: m 2 t – 1 (i +1) 2 m – t – 1 (i +1) 2 m – t – 1 m 2 t – 1 (i +1) 2 m – t – 1 (i +1) 2 m – t – 1 ( w t ( (x j u j ) (y j v j ) ) ) ( w t ( (x j u j ) (y j v j ) ) ) t =0 i =0 j =i 2 m – t j =i 2 m – t variables x 0, … y 0, … u 0, … v 0, … w 0, …. n = 2 m