Local Decoding and Testing Polynomials over Grids

Similar documents
Property Testing and Affine Invariance Part I Madhu Sudan Harvard University

Low-Degree Polynomials

General Strong Polarization

Algebraic Codes and Invariance

Locality in Coding Theory

Two Decades of Property Testing Madhu Sudan Harvard

General Strong Polarization

Uncertain Compression & Graph Coloring. Madhu Sudan Harvard

Reed-Muller testing and approximating small set expansion & hypergraph coloring

Low-Degree Testing. Madhu Sudan MSR. Survey based on many works. of /02/2015 CMSA: Low-degree Testing 1

General Strong Polarization

General Strong Polarization

Two Decades of Property Testing

Fully Understanding the Hashing Trick

Testing Affine-Invariant Properties

Approximate Second Order Algorithms. Seo Taek Kong, Nithin Tangellamudi, Zhikai Guo

Worksheets for GCSE Mathematics. Quadratics. mr-mathematics.com Maths Resources for Teachers. Algebra

Sub-Constant Error Low Degree Test of Almost Linear Size

Invariance in Property Testing

Highly resilient correctors for polynomials

Lecture 8 (Notes) 1. The book Computational Complexity: A Modern Approach by Sanjeev Arora and Boaz Barak;

Support Vector Machines. CSE 4309 Machine Learning Vassilis Athitsos Computer Science and Engineering Department University of Texas at Arlington

7.3 The Jacobi and Gauss-Seidel Iterative Methods

Homomorphism Testing

CS 151 Complexity Theory Spring Solution Set 5

Learning and Fourier Analysis

On the Fourier spectrum of symmetric Boolean functions

Low Rate Is Insufficient for Local Testability

compare to comparison and pointer based sorting, binary trees

Introduction Long transparent proofs The real PCP theorem. Real Number PCPs. Klaus Meer. Brandenburg University of Technology, Cottbus, Germany

A Posteriori Error Estimates For Discontinuous Galerkin Methods Using Non-polynomial Basis Functions

Recent Developments on Circuit Satisfiability Algorithms

Testing by Implicit Learning

Power of Adiabatic Quantum Computation

Crash course Verification of Finite Automata Binary Decision Diagrams

Lecture No. 1 Introduction to Method of Weighted Residuals. Solve the differential equation L (u) = p(x) in V where L is a differential operator

Classical RSA algorithm

International Journal of Mathematical Archive-5(3), 2014, Available online through ISSN

International Journal of Mathematical Archive-5(12), 2014, Available online through ISSN

Definition: A sequence is a function from a subset of the integers (usually either the set

Math 3336 Section 4.3 Primes and Greatest Common Divisors. Prime Numbers and their Properties Conjectures and Open Problems About Primes

Variations. ECE 6540, Lecture 02 Multivariate Random Variables & Linear Algebra

Grover s algorithm. We want to find aa. Search in an unordered database. QC oracle (as usual) Usual trick

: Error Correcting Codes. October 2017 Lecture 1

Algebraic Property Testing: The Role of Invariance

Quantum Chosen-Ciphertext Attacks against Feistel Ciphers

Compressed representation of Kohn-Sham orbitals via selected columns of the density matrix

Symmetric LDPC codes are not necessarily locally testable

Math 95--Review Prealgebra--page 1

Questions Pool. Amnon Ta-Shma and Dean Doron. January 2, Make sure you know how to solve. Do not submit.

The Tensor Product of Two Codes is Not Necessarily Robustly Testable

THE KENNESAW STATE UNIVERSITY HIGH SCHOOL MATHEMATICS COMPETITION PART II Calculators are NOT permitted Time allowed: 2 hours

M.5 Modeling the Effect of Functional Responses

Introduction to Algorithms

Testing Lipschitz Functions on Hypergrid Domains

Testing Low-Degree Polynomials over Prime Fields

Cryptography CS 555. Topic 22: Number Theory/Public Key-Cryptography

Angular Momentum, Electromagnetic Waves

Recurrence Relations

Lecture 6. k+1 n, wherein n =, is defined for a given

Two Query PCP with Sub-Constant Error

PCP Theorem and Hardness of Approximation

Towards the Sliding Scale Conjecture (Old & New PCP constructions)

Predicting Winners of Competitive Events with Topological Data Analysis

Lecture 6. Notes on Linear Algebra. Perceptron

Control of Mobile Robots

Symmetric LDPC codes are not necessarily locally testable

On Axis-Parallel Tests for Tensor Product Codes

Solar Photovoltaics & Energy Systems

Big Bang Planck Era. This theory: cosmological model of the universe that is best supported by several aspects of scientific evidence and observation

B(w, z, v 1, v 2, v 3, A(v 1 ), A(v 2 ), A(v 3 )).

Lecture 22. m n c (k) i,j x i x j = c (k) k=1

Learning and Fourier Analysis

1. The graph of a function f is given above. Answer the question: a. Find the value(s) of x where f is not differentiable. Ans: x = 4, x = 3, x = 2,

Math 3 Unit 3: Polynomial Functions

10.4 The Cross Product

6.895 PCP and Hardness of Approximation MIT, Fall Lecture 3: Coding Theory

Robust local testability of tensor products of LDPC codes

Compatible probability measures

PHY103A: Lecture # 4

Lecture 6. Today we shall use graph entropy to improve the obvious lower bound on good hash functions.

Number Representations

Work, Energy, and Power. Chapter 6 of Essential University Physics, Richard Wolfson, 3 rd Edition

Polynomial Representations of Threshold Functions and Algorithmic Applications. Joint with Josh Alman (Stanford) and Timothy M.

Introduction to Algorithms

2.4 Error Analysis for Iterative Methods

CS60020: Foundations of Algorithm Design and Machine Learning. Sourangshu Bhattacharya

Relaxed Locally Correctable Codes in Computationally Bounded Channels

Course Business. Homework 3 Due Now. Homework 4 Released. Professor Blocki is travelling, but will be back next week

Math 3 Unit 3: Polynomial Functions

Testing Low-Degree Polynomials over GF (2)

With P. Berman and S. Raskhodnikova (STOC 14+). Grigory Yaroslavtsev Warren Center for Network and Data Sciences

Hopfield Network for Associative Memory

Introduction to Algorithms

Math, Stats, and Mathstats Review ECONOMETRICS (ECON 360) BEN VAN KAMMEN, PHD

Worksheets for GCSE Mathematics. Algebraic Expressions. Mr Black 's Maths Resources for Teachers GCSE 1-9. Algebra

Prof. Dr.-Ing. Armin Dekorsy Department of Communications Engineering. Stochastic Processes and Linear Algebra Recap Slides

15-251: Great Theoretical Ideas In Computer Science Recitation 9 : Randomized Algorithms and Communication Complexity Solutions

Lecture 4: Codes based on Concatenation

Higher-order Fourier Analysis over Finite Fields, and Applications. Pooya Hatami

Transcription:

Local Decoding and Testing Polynomials over Grids Madhu Sudan Harvard University Joint work with Srikanth Srinivasan (IIT Bombay) January 11, 2018 ITCS: Polynomials over Grids 1 of 12

DeMillo-Lipton-Schwarz-Zippel Notation: FF = field, FF(dd, nn) = set of deg. dd, nn var. poly over FF δδ ff, gg = normalized Hamming distance δδ dd ff = min gg FF dd,nn δδ(ff, gg) DLSZ Lemma: Let SS FF, with 2 SS <. If ff, gg: SS nn FF satisfy ff gg FF dd, nn then δδ ff, gg 2 dd. Strengthens to δδ ff, gg 1 dd/ SS if dd < SS Holds even if SS = {0,1} For this talk: dd = constant DLSZ Lemma converts polynomials into error-correcting codes. With bonus: decodability, locality so δδ ff, gg bounded away from 0. January 11, 2018 ITCS: Polynomials over Grids 2 of 12

Local Decoding and Testing Informally: Given oracle access to ff: SS nn FF Testing: determine if δδ dd ff εε Decoding: If δδ dd ff δδ 0 and gg is nearest poly then determine value of gg(aa) given aa SS nn Parameters: Query complexity = # of oracle queries Testing accuracy Pr rejection δδ dd ff Decoding distance (δδ 0 above) Ideal: Query complexity OO dd (1), accuracy, dec. distance Ω dd (1) Terminology FF(dd, nn) testable (decodable) over SS if ideal achievable. Thms: For finite FF, FF(dd, nn) decodable over FF (folklore) and testable over FF ([Rubinfeld+S. 96]. [Kaufman+Ron 04]) January 11, 2018 ITCS: Polynomials over Grids 3 of 12

Polynomials over Grids Testing + decoding thms hold only for ff: FF nn FF Not surprising repeated occurrence with polynomials. Even classical decoding algorithms (non-local) work only in this case! [Kopparty-Kim 16]: Non-local Decoder (up to half the distance) for ff: SS nn FF for all finite SS This talk: Testing + decoding when SS = 0,1 grid (Note: equivalent to SS 1 SS nn with SS ii = 2) January 11, 2018 ITCS: Polynomials over Grids 4 of 12

Obstacles to decoding/testing Standard insight behind testing/decoding dd < FF ff has deg. dd iff ff restricted to lines has dd Can test/decode function on lines. Reduces complexity from FF nn to FF. When dd > FF use subspaces (= collection of linear restrictions) Affine invariance 2 transitivity Problem over grids: General lines don t stay within grid! Only linear restrictions that stay in grid: xx ii = 0 or xx ii = 1 xx ii = xx jj or xx ii = 1 xx jj (denoted xx ii = xx jj bb ) January 11, 2018 ITCS: Polynomials over Grids 5 of 12

Main Results Thm 1. FF(1, nn) not locally decodable over grid if char(ff) No 2-transitivity! Serious obstacle! Thm 2. FF(dd, nn) is locally decodable over grid if char FF <. O(max dd,char FF ) Query complexity = 2 Decoding without 2-transitivity! Thm 3. FF(dd, nn) is locally testable over grid ( FF) Testing without decoding! January 11, 2018 ITCS: Polynomials over Grids 6 of 12

Proofs: Not LDC over Q Idea dd = 1 : Consider ff = linear function that is erased on imbalanced points (of weight nn 2 cc nn, nn 2 + cc nn ) Claim 1: No local constraints over Q on balanced points and 0 nn Claim 1.1: if ff 0 nn = l ii=1 aa ii ff xx ii ff then 0 nn = l ii=1 aa ii xx ii Claim 1.2: 1 aa ii l! Claim 1.3: If xx 1,, xx l are εε-balanced jj jj then ii aa ii xx ii is εε ii aa ii Claim 2: No local constraints Not LDC Known over finite fields [BHR 04] Not immediate for Q -balanced January 11, 2018 ITCS: Polynomials over Grids 7 of 12

Proofs: LDC over small characteristic Given: ff δδ-close to gg FF(dd, nn), aa 0,1 nn (char FF = pp) Pick σσ: nn kk uniformly and let xx ii = aa ii yy σσ(ii) So considering ff yy 1,, yy kk = ff aa yy σσ : ff 0 = ff aa ; gg 0 = gg(aa) Claim 1: For balanced yy, have aa yy σσ aa Claim 2: If kk 2 = pptt and kk 2 > dd then gg (0) determined by gg (yy) yy = kk 2 (Usual ingredient Lucas s Theorem xx pp + yy pp = xx + yy pp ) January 11, 2018 ITCS: Polynomials over Grids 8 of 12

Proofs: LTC over all fields: Test Test Pick aa 0,1 nn uniformly and σσ: nn kk randomly Verify ff yy 1,, yy kk = ff aa yy σσ is of degree dd Randomly =? Not uniformly (don t know how to analyze). Instead random union-find Initially each xx ii in iith bucket. Iterate: pick two uniformly random buckets and merge till kk buckets left. Assign yy 1,, yy kk to kk buckets randomly. Allows for proof by induction January 11, 2018 ITCS: Polynomials over Grids 9 of 12

Proofs: LTC over all field: Analysis Key ingredients: (mimics BKSSZ analysis for RM) Main claim: Pr test rejects ff 2 OO(dd) min 1, δδ dd (ff) Proof by induction on nn with 3 cases: Case 1: ff moderately close to deg. dd Show that eventually ff disagrees from nearby poly on exactly one of 2 kk samples. Usual proof pairwise independence Our proof hypercontractivity of sphere [Polyanskiy] + prob. fact about random union-find Case 2: Pr ii,jj,bb ff xx ii =xx jj bb very close to deg. dd small Induction Case 3: ff far from deg. dd, but Pr ii,jj,bb high Use algebra to get contradiction. (Stitch different polynomials from diff. (ii, jj, bb) s to get nearby poly to ff) January 11, 2018 ITCS: Polynomials over Grids 10 of 12

Conclusions + Open questions Many aspects of polynomials well-understood only over domain = FF nn Grid setting (SS nn ) far less understood. When SS = {0,1} (equivalently SS = 2) testing possible without local decoding! (novel in context of low-degree tests!) General SS open!! (even SS = 3 or even SS = 0,1,2?) Is there a gap between characterizability and testability here? January 11, 2018 ITCS: Polynomials over Grids 11 of 12

Thank You! January 11, 2018 ITCS: Polynomials over Grids 12 of 12