On Derandomizing Algorithms that Err Extremely Rarely

Slides:



Advertisements
Similar presentations
How to Solve Longstanding Open Problems In Quantum Computing Using Only Fourier Analysis Scott Aaronson (MIT) For those who hate quantum: The open problems.
Advertisements

On the Complexity of Parallel Hardness Amplification for One-Way Functions Chi-Jen Lu Academia Sinica, Taiwan.
Finding Cycles and Trees in Sublinear Time Oded Goldreich Weizmann Institute of Science Joint work with Artur Czumaj, Dana Ron, C. Seshadhri, Asaf Shapira,
Pseudorandomness from Shrinkage David Zuckerman University of Texas at Austin Joint with Russell Impagliazzo and Raghu Meka.
Linear-Degree Extractors and the Inapproximability of Max Clique and Chromatic Number David Zuckerman University of Texas at Austin.
Models of Computation Prepared by John Reif, Ph.D. Distinguished Professor of Computer Science Duke University Analysis of Algorithms Week 1, Lecture 2.
Extracting Randomness From Few Independent Sources Boaz Barak, IAS Russell Impagliazzo, UCSD Avi Wigderson, IAS.
Computational Analogues of Entropy Boaz Barak Ronen Shaltiel Avi Wigderson.
Boolean Circuits of Depth-Three and Arithmetic Circuits with General Gates Oded Goldreich Weizmann Institute of Science Based on Joint work with Avi Wigderson.
Pseudorandomness from Shrinkage David Zuckerman University of Texas at Austin Joint with Russell Impagliazzo and Raghu Meka.
Approximate List- Decoding and Hardness Amplification Valentine Kabanets (SFU) joint work with Russell Impagliazzo and Ragesh Jaiswal (UCSD)
Talk for Topics course. Pseudo-Random Generators pseudo-random bits PRG seed Use a short “ seed ” of very few truly random bits to generate a long string.
Simple extractors for all min- entropies and a new pseudo- random generator Ronen Shaltiel Chris Umans.
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.
DNF Sparsification and Counting Raghu Meka (IAS, Princeton) Parikshit Gopalan (MSR, SVC) Omer Reingold (MSR, SVC)
Circuit Complexity and Derandomization Tokyo Institute of Technology Akinori Kawachi.
Deterministic Amplification of Space-Bounded Probabilistic Algorithms Ziv Bar-Yossef Oded Goldreich U.C. Berkeley Weizmann Institute U.C. Berkeley Weizmann.
Better Pseudorandom Generators from Milder Pseudorandom Restrictions Raghu Meka (IAS) Parikshit Gopalan, Omer Reingold (MSR-SVC) Luca Trevian (Stanford),
Derandomized parallel repetition theorems for free games Ronen Shaltiel, University of Haifa.
Using Nondeterminism to Amplify Hardness Emanuele Viola Joint work with: Alex Healy and Salil Vadhan Harvard University.
Some Thoughts regarding Unconditional Derandomization Oded Goldreich Weizmann Institute of Science RANDOM 2010.
Derandomization: New Results and Applications Emanuele Viola Harvard University March 2006.
On Uniform Amplification of Hardness in NP Luca Trevisan STOC 05 Paper Review Present by Hai Xu.
Arithmetic Hardness vs. Randomness Valentine Kabanets SFU.
Derandomizing LOGSPACE Based on a paper by Russell Impagliazo, Noam Nissan and Avi Wigderson Presented by Amir Rosenfeld.
1 Streaming Computation of Combinatorial Objects Ziv Bar-Yossef U.C. Berkeley Omer Reingold AT&T Labs – Research Ronen.
In a World of BPP=P Oded Goldreich Weizmann Institute of Science.
Ragesh Jaiswal Indian Institute of Technology Delhi Threshold Direct Product Theorems: a survey.
Why Extractors? … Extractors, and the closely related “Dispersers”, exhibit some of the most “random-like” properties of explicitly constructed combinatorial.
On Constructing Parallel Pseudorandom Generators from One-Way Functions Emanuele Viola Harvard University June 2005.
Direct-product testing, and a new 2-query PCP Russell Impagliazzo (IAS & UCSD) Valentine Kabanets (SFU) Avi Wigderson (IAS)
XOR lemmas & Direct Product thms - Many proofs Avi Wigderson IAS, Princeton ’82 Yao ’87 Levin ‘89 Goldreich-Levin ’95 Impagliazzo ‘95 Goldreich-Nisan-Wigderson.
Using Nondeterminism to Amplify Hardness Emanuele Viola Joint work with: Alex Healy and Salil Vadhan Harvard University.
1 New Coins from old: Computing with unknown bias Elchanan Mossel, U.C. Berkeley
Circuit Lower Bounds via Ehrenfeucht- Fraïssé Games Michal Koucký Joint work with: Clemens Lautemann, Sebastian Poloczek, Denis Thérien.
Amplification and Derandomization Without Slowdown Dana Moshkovitz MIT Joint work with Ofer Grossman (MIT)
Umans Complexity Theory Lectures Lecture 17: Natural Proofs.
Norms, XOR lemmas, and lower bounds for GF(2) polynomials and multiparty protocols Emanuele Viola, IAS (Work partially done during postdoc at Harvard)
RANDOMNESS VS. MEMORY: Prospects and Barriers Omer Reingold, Microsoft Research and Weizmann With insights courtesy of Moni Naor, Ran Raz, Luca Trevisan,
Pseudorandom Bits for Constant-Depth Circuits with Few Arbitrary Symmetric Gates Emanuele Viola Harvard University June 2005.
Comparing Notions of Full Derandomization Lance Fortnow NEC Research Institute With thanks to Dieter van Melkebeek.
Almost SL=L, and Near-Perfect Derandomization Oded Goldreich The Weizmann Institute Avi Wigderson IAS, Princeton Hebrew University.
Umans Complexity Theory Lecturess Lecture 11: Randomness Extractors.
Umans Complexity Theory Lectures Lecture 9b: Pseudo-Random Generators (PRGs) for BPP: - Hardness vs. randomness - Nisan-Wigderson (NW) Pseudo- Random Generator.
Randomness and Computation
Introduction to Quantum Computing Lecture 1 of 2
Finding Cycles and Trees in Sublinear Time
A new characterization of ACC0 and probabilistic CC0
Circuit Lower Bounds A combinatorial approach to P vs NP
Modern symmetric-key Encryption
Pseudorandomness when the odds are against you
Background: Lattices and the Learning-with-Errors problem
Secure Computation of Constant-Depth Circuits with Applications to Database Search Problems Omer Barkol Yuval Ishai Technion.
Tight Fourier Tails for AC0 Circuits
Pseudo-derandomizing learning and approximation
The Curve Merger (Dvir & Widgerson, 2008)
Non-Malleable Extractors New tools and improved constructions
DNF Sparsification and Counting
Indistinguishability by adaptive procedures with advice, and lower bounds on hardness amplification proofs Aryeh Grinberg, U. Haifa Ronen.
Every set in P is strongly testable under a suitable encoding
Emanuele Viola Harvard University June 2005
CS151 Complexity Theory Lecture 10 May 2, 2019.
Switching Lemmas and Proof Complexity
Oracle Separation of BQP and PH
Stronger Connections Between Circuit Analysis and Circuit Lower Bounds, via PCPs of Proximity Lijie Chen Ryan Williams.
Oracle Separation of BQP and PH
On Probabilistic Time versus Alternating Time
Emanuele Viola Harvard University October 2005
Pseudorandomness: New Results and Applications
Presentation transcript:

On Derandomizing Algorithms that Err Extremely Rarely Oded Goldreich Weizmann Institute of Science Note difference in titles. Based on Joint work with Avi Wigderson

Standard Derandomization Challenges Given a circuit C (from a certain class) such that Prob[C(x)=1] > ½, find an input x such that C(x)=1. Famous frontier: Solve it in poly-time for the class AC0. A two-sided error version: Given a circuit C (from a certain class) such that Prob[C(x)=] > ⅔, for some , find an input x such that C(x)= . Ignore the other versions A black-box version: Given circuit parameters, find a set of inputs S such that every circuit C that satisfies Prob[C(x)=1] > ½, there exists xS such that C(x)=1 .

Quantified Derandomization Challenges (new) For a class C and a bound B, given an n-bit input circuit C from C such that |x:C(x)=0}| < B(n), find an input x such that C(x)=1. The above is called the (C,B)-search problem. Focus: Small B; e.g., quasi-polynomial, sub-exponential. Q: Is the (P/poly,nlog n)-search problem solvable in (deter.) poly-time? Ignore the other versions THM1: The (AC0,exp(n0.999))-search problem is solvable in (deter.) poly-time. THM2: Standard derandomization of AC0 is reducible to the (AC0,exp(n/log n))-search problem.

On Quantified Derandomization Problems (C,B)-search problem = given an n-bit input circuit C from C such that |x:C(x)=0}| < B(n), find an input x such that C(x)=1. THM1: The (AC0,exp(n0.999))-search problem is solvable in (deter.) poly-time. THM2: Standard derandomization of AC0 is reducible to the (AC0,exp(n/log n))-search problem. AC0[2] = constant-depth poly-size circuits with (unbounded fan-in) parity gates (in addition to the standard unbounded AND and OR gates) THM3: Standard derandomization of AC0[2] is reducible to the (AC0[2],exp(n0.001))-search problem.

On the proof of THM 1 (C,B)-search problem = given an n-bit input circuit C from C such that |x:C(x)=0}| < B(n), find an input x such that C(x)=1. THM1: The (AC0,exp(n0.999))-search problem is solvable in (deter.) poly-time. Idea: Hit the circuit with a pseudorandom restriction (generated based on a seed of log length) such that (i) at least 2n0.999 variables survive, and (ii) the circuit simplifies to a constant. The PR-restriction may not preserve the acceptance probability of a generic AC0 circuit (not even approximately), but this suffices for us since the number of surviving variables exceeds the error bound B(n). When designing the PR-restriction we focus on the simplification, which is obtained by repeated applications of a PR switching lemma.

On the proof of THM 3 (C,B)-search problem = given an n-bit input circuit C from C such that |x:C(x)=0}| < B(n), find an input x such that C(x)=1. THM3: Standard derandomization of AC0[2] is reducible to the (AC0[2],exp(n0.001))-search problem. Idea: Given a circuit for the standard problem, obtain a circuit for the quantified problem via “radical” error reduction (using a randomness extractor). Extractors for min-entropy n0.001 that use a seed of log length and extract n0.0005 bits can be computed in AC0[2], whereas approximate majority can be computed in AC0. C’(x) = invoke C(E(x,s)) on all poly-many seeds, and take approx.-majority. The “bad” inputs yields a flat source that violates extraction.

On the proof of THM 2 (C,B)-search problem = given an n-bit input circuit C from C such that |x:C(x)=0}| < B(n), find an input x such that C(x)=1. THM2: Standard derandomization of AC0 is reducible to the (AC0,exp(n/log n))-search problem. We cannot perform error-reduction in AC0 since we do not know of an adequate extractor computable in AC0. But it suffices to perform error-reduction for AC0! (i) Reduce randomness to polylog (via PRG), (ii) perform radical error reduction (via an extractor), (iii) straightforward error+randomness amplification. We need to extract nOmega(1) bits from a source of (sub)constant entropy rate. It suffices that the extractor’s output fools AC0. “in a class” = via a randomness extractor, whereas “for a class” allows using a “pseudo-extractor” that only fools the class.

Summary: Quantified Derandomization (C,B)-search problem = given an n-bit input circuit C from C such that |x:C(x)=0}| < B(n), find an input x such that C(x)=1. Q: Is the (P/poly,nlog n)-search problem solvable in (deter.) poly-time? Results regarding quantified derand’ may improve over the known for standard derand’ (see Thm1), but in some cases quantified derand’ implies standard derand’ (see Thm2 and Thm3). Hope: A smooth transition. E.g., for AC0, from B(n)=exp(n0.999) as in Thm1 to B(n)=exp(n/log n) as in Thm2. Mention results regarding MA0 and AM0=AM (which preserves error)?

END Slides available at http://www.wisdom.weizmann.ac.il/~oded/T/aq.pptx Paper available at http://www.wisdom.weizmann.ac.il/~oded/p_aq.html