NP-Completeness

Total Page:16

File Type:pdf, Size:1020Kb

NP-Completeness NPcompleteness A Retrosp ective Christos H Papadimitriou University of California Berkeley USA Abstract For a quarter of a century now NPcompleteness has b een computer sciences favorite paradigm fad punching bag buzzword alibi and intellectual exp ort This pap er is a fragmentary commentary on its origins its nature its impact and on the attributes that have made it so p ervasive and contagious A keyword search in Melvyl the University of Californias online library reveals that ab out pap ers each year have the term NPcomplete on their title abstract or list of keywords This is more than each of the terms compiler database exp ert neural network and op erating system Even more surprising is the diversity of the disciplines with pap ers referring to NPcompleteness They range from statistics and articial life to automatic control and nuclear engineering What is the nature and extent of the impact of NPcompleteness on theoretical computer science computer science in general computing practice as well as other domains of the natural sciences applied sci ence and mathematics And why did NPcompleteness b ecome such a p ervasive and inuential concept One of the reasons of the immense impact of NPcompleteness has to b e the app eal and elegance of the class P that is of the thesis that p olynomial worstcase time is a plausible and pro ductive mathematical surrogate of the empirical concept of practically solvable computational problem But obvi ously NPcompleteness also draws on the imp ortance of NP as it rests on the widely conjectured contradistinction b etween these two classes In this regard it is crucial that NP captures vast domains of computational scientic and mathematical endeavor and seems to roughly delimit what mathematicians and scientists had b een aspiring to compute feasibly True there are domains such as strategic analysis and counting which have b een within our computational ambitions and still seem to lie outside NP but they are the exceptions rather than the rule NPcompleteness has thus b ecome a valuable intermediary b e tween the abstraction of computational mo dels and the reality of computational problems grounding complexity theory to computational practice Also crucial for the success of NPcompleteness has b een its surprising ubiq uity and eectiveness as a classication to ol and the scarcity of problems in ? christoscsberkeleyedu Partially supp orted by the National Science Foundation A version of this talk was given at a meeting in the Fall of celebrating the th birthday of Richard M Karp to whom this pap er is also aectionately dedicated NP that resist classication as either p olynomialtime solvable or NPcomplete Ladners result on intermediate degrees b etween P and NPcompleteness had b een known almost as so on as NPcompleteness was introduced and thus theoretically the world could b e full of mysterious intermediate problems In sev eral o ccasions extremely broad classes of computational problems in NP have b een dichotomized with surprising accuracy into p olynomially solvable and NP complete see for two early examples The founders of NPcompleteness app ear to have anticipated its broad applicability and classication p ower Leonid Levin wrote in The method described here clearly provides a means for readily obtaining re sults of this type for the majority of important sequential search problems In Karps pap er twenty one problems were proved NPcomplete showing b e yond any doubt the surprisingly broad applicability of the metho d Signicantly Karp seems annoyed and surprised that three other problems linear program ming primality and graph isomorphism resisted at the time such classication Primality and graph isomorphism were also mentioned by Co ok Knuth was suciently convinced ab out the imp ortance and broad applicability of the new concept to take early and delib erate action on the terminological front NPcompleteness has had tremendous impact even in areas where in some sense it should not have It is now common knowledge among computer sci entists that NPcompleteness is largely irrelevant to publickey cryptography since in that area one needs sophisticated cryptographic assumptions that go b eyond NPcompleteness and worstcase p olynomialtime computation fur thermore cryptographic proto cols based on NPcomplete problems have b een illfated Fortunately the founders of mo dern cryptography did not know this Die and Hellman base their famous pronouncement We stand today on the brink of a revolution in cryptography on two facts Very fast hardware and software and novel techniques for proving problems hard they cite Karps pap er NPcompleteness has also exhibited a great amount of versatility adapting to contexts and computational asp ects b eyond its original scop e of worstcase analysis of exact algorithms for decision and optimization problems For exam ple it was used early on to show that certain optimization problems cannot b e approximated satisfactorily and indeed in a most ingenious and compre hensive way more recently By showing that even less ambitious goals than worstcase p olynomial exact solution are unattainable NPcompleteness is thus a most useful to ol for rep eatedly pruning unpromising research directions and thus redirecting research to new ones in a manner reminiscent of the struggle b etween Hercules and the monster Hydra Let me illustrate this versatility of NPcompleteness by a technical interlude on an asp ect of ecient computation that has interested me recently namely output polynomial time Certain computational problems require an output f x on input x that is in the worst case exp onential in the input For such problems one would like to have algorithms that are p olynomial in jxj and jf xj The class of problems thus solvable can b e called output polynomial time One can use NP completeness to prove that certain functions are not in outputp olynomial time unless PNP For example consider the function MIN which maps a regular expression to the minimumstate equivalent deterministic nitestate automaton MIN can b e computed by rst designing a nondeterministic automaton M then 0 an equivalent deterministic automaton M and next minimizing the states of 0 M to obtain the nal output the problem is of course that the intermediate 0 result M could b e exp onential in both the input and the output It is rather straightforward to use traditional NPcompleteness techniques to show the following Theorem Unless PNP MIN is not in output polynomial time In fact we cannot even compute in outputp olynomial time a deterministic au tomaton that has at most polynomially more states than the minimum unless of course PNP Often the required output f x is a set fy y g of strings that are related k to x via an NP mapping for example if G is a graph let AMISG b e the set of all maximal indep endent sets of G AMIS is known to b e in outputp olynomial time see for an exp osition and strengthening of this result and an early discussion of output p olynomial time For such problems we have an elegant is alternative denition of output p olynomial time A function f in output p olynomial time if the following problem is solvable in p olynomial time Given x and y either decide that y f x or nd a string in y f x It is easy to see that if such an algorithm exists then its iteration starting with S gives an output p olynomial time algorithm for f and vice versa if an output p olynomial time algorithm exists for f it can b e used to pro duce an element of y f x For example AMIS is in output p olynomial time its generalization to hypergraphs is op en but was recently shown to b e c log n in output n time see for an extensive discussion of the hypergraph generalization of AMIS One can use again traditional NPcompleteness to show that the following generalization is not in output p olynomial time unless PNP Given a monotone circuit compute the set of all minimal with resp ect to the set of true inputs satisfying truth assignments But sometimes traditional NPcompleteness techniques do not seem to suf ce to bring out the intractability of a problem b ecause this problem b elongs to a class or computational mo de that app ears to b e b etween P and NP In such cases NPcompleteness has acted as an op enended research paradigm spawn ing variants that are appropriate for the computational context b eing studied examples are classes that capture lo cal search the parity argument loga rithmic nondeterminism the related concept of xedparameter tractability and approximability Complexity classes introduced this way as abstractions of natural compu tational problems of mysteriously intermediate complexity are in some precise sense wellmotivated indeed necessary they are discovered not invented as they have always existed by dint of their natural complete problems The only way to make them go away is to collapse them with P or NP as o ccasionally happ ens recall and its brilliant followup NPcompleteness is of course a valuable to ol for demonstrating the diculty of computational problems However NPcompleteness is often used allegori cally a problem is shown NPcomplete that is not strictly sp eaking a natural computational problem but an articial problem created to capture a mathe matical concept NPcompleteness in this context suggests that a problem area or approach is mathematically nasty Because if we b elieve that ecient algo rithms are the natural outow of the mathematical structure of a problem a view shared by all computer scientists with the p ossible
Recommended publications
  • On the Randomness Complexity of Interactive Proofs and Statistical Zero-Knowledge Proofs*
    On the Randomness Complexity of Interactive Proofs and Statistical Zero-Knowledge Proofs* Benny Applebaum† Eyal Golombek* Abstract We study the randomness complexity of interactive proofs and zero-knowledge proofs. In particular, we ask whether it is possible to reduce the randomness complexity, R, of the verifier to be comparable with the number of bits, CV , that the verifier sends during the interaction. We show that such randomness sparsification is possible in several settings. Specifically, unconditional sparsification can be obtained in the non-uniform setting (where the verifier is modelled as a circuit), and in the uniform setting where the parties have access to a (reusable) common-random-string (CRS). We further show that constant-round uniform protocols can be sparsified without a CRS under a plausible worst-case complexity-theoretic assumption that was used previously in the context of derandomization. All the above sparsification results preserve statistical-zero knowledge provided that this property holds against a cheating verifier. We further show that randomness sparsification can be applied to honest-verifier statistical zero-knowledge (HVSZK) proofs at the expense of increasing the communica- tion from the prover by R−F bits, or, in the case of honest-verifier perfect zero-knowledge (HVPZK) by slowing down the simulation by a factor of 2R−F . Here F is a new measure of accessible bit complexity of an HVZK proof system that ranges from 0 to R, where a maximal grade of R is achieved when zero- knowledge holds against a “semi-malicious” verifier that maliciously selects its random tape and then plays honestly.
    [Show full text]
  • Randomised Computation 1 TM Taking Advices 2 Karp-Lipton Theorem
    INFR11102: Computational Complexity 29/10/2019 Lecture 13: More on circuit models; Randomised Computation Lecturer: Heng Guo 1 TM taking advices An alternative way to characterize P=poly is via TMs that take advices. Definition 1. For functions F : N ! N and A : N ! N, the complexity class DTime[F ]=A consists of languages L such that there exist a TM with time bound F (n) and a sequence fangn2N of “advices” satisfying: • janj ≤ A(n); • for jxj = n, x 2 L if and only if M(x; an) = 1. The following theorem explains the notation P=poly, namely “polynomial-time with poly- nomial advice”. S c Theorem 1. P=poly = c;d2N DTime[n ]=nd . Proof. If L 2 P=poly, then it can be computed by a family C = fC1;C2; · · · g of Boolean circuits. Let an be the description of Cn, andS the polynomial time machine M just reads 2 c this description and simulates it. Hence L c;d2N DTime[n ]=nd . For the other direction, if a language L can be computed in polynomial-time with poly- nomial advice, say by TM M with advices fang, then we can construct circuits fDng to simulate M, as in the theorem P ⊂ P=poly in the last lecture. Hence, Dn(x; an) = 1 if and only if x 2 L. The final circuit Cn just does exactly what Dn does, except that Cn “hardwires” the advice an. Namely, Cn(x) := Dn(x; an). Hence, L 2 P=poly. 2 Karp-Lipton Theorem Dick Karp and Dick Lipton showed that NP is unlikely to be contained in P=poly [KL80].
    [Show full text]
  • Computational Complexity and Intractability: an Introduction to the Theory of NP Chapter 9 2 Objectives
    1 Computational Complexity and Intractability: An Introduction to the Theory of NP Chapter 9 2 Objectives . Classify problems as tractable or intractable . Define decision problems . Define the class P . Define nondeterministic algorithms . Define the class NP . Define polynomial transformations . Define the class of NP-Complete 3 Input Size and Time Complexity . Time complexity of algorithms: . Polynomial time (efficient) vs. Exponential time (inefficient) f(n) n = 10 30 50 n 0.00001 sec 0.00003 sec 0.00005 sec n5 0.1 sec 24.3 sec 5.2 mins 2n 0.001 sec 17.9 mins 35.7 yrs 4 “Hard” and “Easy” Problems . “Easy” problems can be solved by polynomial time algorithms . Searching problem, sorting, Dijkstra’s algorithm, matrix multiplication, all pairs shortest path . “Hard” problems cannot be solved by polynomial time algorithms . 0/1 knapsack, traveling salesman . Sometimes the dividing line between “easy” and “hard” problems is a fine one. For example, . Find the shortest path in a graph from X to Y (easy) . Find the longest path (with no cycles) in a graph from X to Y (hard) 5 “Hard” and “Easy” Problems . Motivation: is it possible to efficiently solve “hard” problems? Efficiently solve means polynomial time solutions. Some problems have been proved that no efficient algorithms for them. For example, print all permutation of a number n. However, many problems we cannot prove there exists no efficient algorithms, and at the same time, we cannot find one either. 6 Traveling Salesperson Problem . No algorithm has ever been developed with a Worst-case time complexity better than exponential .
    [Show full text]
  • Valuative Characterization of Central Extensions of Algebraic Tori on Krull
    Valuative characterization of central extensions of algebraic tori on Krull domains ∗ Haruhisa Nakajima † Department of Mathematics, J. F. Oberlin University Tokiwa-machi, Machida, Tokyo 194-0294, JAPAN Abstract Let G be an affine algebraic group with an algebraic torus G0 over an alge- braically closed field K of an arbitrary characteristic p. We show a criterion for G to be a finite central extension of G0 in terms of invariant theory of all regular 0 actions of any closed subgroup H containing ZG(G ) on affine Krull K-schemes such that invariant rational functions are locally fractions of invariant regular func- tions. Consider an affine Krull H-scheme X = Spec(R) and a prime ideal P of R with ht(P) = ht(P ∩ RH ) = 1. Let I(P) denote the inertia group of P un- der the action of H. The group G is central over G0 if and only if the fraction e(P, P ∩ RH )/e(P, P ∩ RI(P)) of ramification indices is equal to 1 (p = 0) or to the p-part of the order of the group of weights of G0 on RI(P)) vanishing on RI(P))/P ∩ RI(P) (p> 0) for an arbitrary X and P. MSC: primary 13A50, 14R20, 20G05; secondary 14L30, 14M25 Keywords: Krull domain; ramification index; algebraic group; algebraic torus; char- acter group; invariant theory 1 Introduction 1.A. We consider affine algebraic groups and affine schemes over a fixed algebraically closed field K of an arbitrary characteristic p. For an affine group G, denote by G0 its identity component.
    [Show full text]
  • Computational Complexity: a Modern Approach
    i Computational Complexity: A Modern Approach Draft of a book: Dated January 2007 Comments welcome! Sanjeev Arora and Boaz Barak Princeton University [email protected] Not to be reproduced or distributed without the authors’ permission This is an Internet draft. Some chapters are more finished than others. References and attributions are very preliminary and we apologize in advance for any omissions (but hope you will nevertheless point them out to us). Please send us bugs, typos, missing references or general comments to [email protected] — Thank You!! DRAFT ii DRAFT Chapter 9 Complexity of counting “It is an empirical fact that for many combinatorial problems the detection of the existence of a solution is easy, yet no computationally efficient method is known for counting their number.... for a variety of problems this phenomenon can be explained.” L. Valiant 1979 The class NP captures the difficulty of finding certificates. However, in many contexts, one is interested not just in a single certificate, but actually counting the number of certificates. This chapter studies #P, (pronounced “sharp p”), a complexity class that captures this notion. Counting problems arise in diverse fields, often in situations having to do with estimations of probability. Examples include statistical estimation, statistical physics, network design, and more. Counting problems are also studied in a field of mathematics called enumerative combinatorics, which tries to obtain closed-form mathematical expressions for counting problems. To give an example, in the 19th century Kirchoff showed how to count the number of spanning trees in a graph using a simple determinant computation. Results in this chapter will show that for many natural counting problems, such efficiently computable expressions are unlikely to exist.
    [Show full text]
  • NP-Completeness Part I
    NP-Completeness Part I Outline for Today ● Recap from Last Time ● Welcome back from break! Let's make sure we're all on the same page here. ● Polynomial-Time Reducibility ● Connecting problems together. ● NP-Completeness ● What are the hardest problems in NP? ● The Cook-Levin Theorem ● A concrete NP-complete problem. Recap from Last Time The Limits of Computability EQTM EQTM co-RE R RE LD LD ADD HALT ATM HALT ATM 0*1* The Limits of Efficient Computation P NP R P and NP Refresher ● The class P consists of all problems solvable in deterministic polynomial time. ● The class NP consists of all problems solvable in nondeterministic polynomial time. ● Equivalently, NP consists of all problems for which there is a deterministic, polynomial-time verifier for the problem. Reducibility Maximum Matching ● Given an undirected graph G, a matching in G is a set of edges such that no two edges share an endpoint. ● A maximum matching is a matching with the largest number of edges. AA maximummaximum matching.matching. Maximum Matching ● Jack Edmonds' paper “Paths, Trees, and Flowers” gives a polynomial-time algorithm for finding maximum matchings. ● (This is the same Edmonds as in “Cobham- Edmonds Thesis.) ● Using this fact, what other problems can we solve? Domino Tiling Domino Tiling Solving Domino Tiling Solving Domino Tiling Solving Domino Tiling Solving Domino Tiling Solving Domino Tiling Solving Domino Tiling Solving Domino Tiling Solving Domino Tiling The Setup ● To determine whether you can place at least k dominoes on a crossword grid, do the following: ● Convert the grid into a graph: each empty cell is a node, and any two adjacent empty cells have an edge between them.
    [Show full text]
  • The Complexity Zoo
    The Complexity Zoo Scott Aaronson www.ScottAaronson.com LATEX Translation by Chris Bourke [email protected] 417 classes and counting 1 Contents 1 About This Document 3 2 Introductory Essay 4 2.1 Recommended Further Reading ......................... 4 2.2 Other Theory Compendia ............................ 5 2.3 Errors? ....................................... 5 3 Pronunciation Guide 6 4 Complexity Classes 10 5 Special Zoo Exhibit: Classes of Quantum States and Probability Distribu- tions 110 6 Acknowledgements 116 7 Bibliography 117 2 1 About This Document What is this? Well its a PDF version of the website www.ComplexityZoo.com typeset in LATEX using the complexity package. Well, what’s that? The original Complexity Zoo is a website created by Scott Aaronson which contains a (more or less) comprehensive list of Complexity Classes studied in the area of theoretical computer science known as Computa- tional Complexity. I took on the (mostly painless, thank god for regular expressions) task of translating the Zoo’s HTML code to LATEX for two reasons. First, as a regular Zoo patron, I thought, “what better way to honor such an endeavor than to spruce up the cages a bit and typeset them all in beautiful LATEX.” Second, I thought it would be a perfect project to develop complexity, a LATEX pack- age I’ve created that defines commands to typeset (almost) all of the complexity classes you’ll find here (along with some handy options that allow you to conveniently change the fonts with a single option parameters). To get the package, visit my own home page at http://www.cse.unl.edu/~cbourke/.
    [Show full text]
  • CS286.2 Lectures 5-6: Introduction to Hamiltonian Complexity, QMA-Completeness of the Local Hamiltonian Problem
    CS286.2 Lectures 5-6: Introduction to Hamiltonian Complexity, QMA-completeness of the Local Hamiltonian problem Scribe: Jenish C. Mehta The Complexity Class BQP The complexity class BQP is the quantum analog of the class BPP. It consists of all languages that can be decided in quantum polynomial time. More formally, Definition 1. A language L 2 BQP if there exists a classical polynomial time algorithm A that ∗ maps inputs x 2 f0, 1g to quantum circuits Cx on n = poly(jxj) qubits, where the circuit is considered a sequence of unitary operators each on 2 qubits, i.e Cx = UTUT−1...U1 where each 2 2 Ui 2 L C ⊗ C , such that: 2 i. Completeness: x 2 L ) Pr(Cx accepts j0ni) ≥ 3 1 ii. Soundness: x 62 L ) Pr(Cx accepts j0ni) ≤ 3 We say that the circuit “Cx accepts jyi” if the first output qubit measured in Cxjyi is 0. More j0i specifically, letting P1 = j0ih0j1 be the projection of the first qubit on state j0i, j0i 2 Pr(Cx accepts jyi) =k (P1 ⊗ In−1)Cxjyi k2 The Complexity Class QMA The complexity class QMA (or BQNP, as Kitaev originally named it) is the quantum analog of the class NP. More formally, Definition 2. A language L 2 QMA if there exists a classical polynomial time algorithm A that ∗ maps inputs x 2 f0, 1g to quantum circuits Cx on n + q = poly(jxj) qubits, such that: 2q i. Completeness: x 2 L ) 9jyi 2 C , kjyik2 = 1, such that Pr(Cx accepts j0ni ⊗ 2 jyi) ≥ 3 2q 1 ii.
    [Show full text]
  • Lecture 11 — October 16, 2012 1 Overview 2
    6.841: Advanced Complexity Theory Fall 2012 Lecture 11 | October 16, 2012 Prof. Dana Moshkovitz Scribe: Hyuk Jun Kweon 1 Overview In the previous lecture, there was a question about whether or not true randomness exists in the universe. This question is still seriously debated by many physicists and philosophers. However, even if true randomness does not exist, it is possible to use pseudorandomness for practical purposes. In any case, randomization is a powerful tool in algorithms design, sometimes leading to more elegant or faster ways of solving problems than known deterministic methods. This naturally leads one to wonder whether randomization can solve problems outside of P, de- terministic polynomial time. One way of viewing randomized algorithms is that for each possible setting of the random bits used, a different strategy is pursued by the algorithm. A property of randomized algorithms that succeed with bounded error is that the vast majority of strategies suc- ceed. With error reduction, all but exponentially many strategies will succeed. A major theme of theoretical computer science over the last 20 years has been to turn algorithms with this property into algorithms that deterministically find a winning strategy. From the outset, this seems like a daunting task { without randomness, how can one efficiently search for such strategies? As we'll see in later lectures, this is indeed possible, modulo some hardness assumptions. There are two main objectives in this lecture. The first objective is to introduce some probabilis- tic complexity classes, by bounding time or space. These classes are probabilistic analogues of deterministic complexity classes P and L.
    [Show full text]
  • Complexity Theory
    Complexity Theory Course Notes Sebastiaan A. Terwijn Radboud University Nijmegen Department of Mathematics P.O. Box 9010 6500 GL Nijmegen the Netherlands [email protected] Copyright c 2010 by Sebastiaan A. Terwijn Version: December 2017 ii Contents 1 Introduction 1 1.1 Complexity theory . .1 1.2 Preliminaries . .1 1.3 Turing machines . .2 1.4 Big O and small o .........................3 1.5 Logic . .3 1.6 Number theory . .4 1.7 Exercises . .5 2 Basics 6 2.1 Time and space bounds . .6 2.2 Inclusions between classes . .7 2.3 Hierarchy theorems . .8 2.4 Central complexity classes . 10 2.5 Problems from logic, algebra, and graph theory . 11 2.6 The Immerman-Szelepcs´enyi Theorem . 12 2.7 Exercises . 14 3 Reductions and completeness 16 3.1 Many-one reductions . 16 3.2 NP-complete problems . 18 3.3 More decision problems from logic . 19 3.4 Completeness of Hamilton path and TSP . 22 3.5 Exercises . 24 4 Relativized computation and the polynomial hierarchy 27 4.1 Relativized computation . 27 4.2 The Polynomial Hierarchy . 28 4.3 Relativization . 31 4.4 Exercises . 32 iii 5 Diagonalization 34 5.1 The Halting Problem . 34 5.2 Intermediate sets . 34 5.3 Oracle separations . 36 5.4 Many-one versus Turing reductions . 38 5.5 Sparse sets . 38 5.6 The Gap Theorem . 40 5.7 The Speed-Up Theorem . 41 5.8 Exercises . 43 6 Randomized computation 45 6.1 Probabilistic classes . 45 6.2 More about BPP . 48 6.3 The classes RP and ZPP .
    [Show full text]
  • R Z- /Li ,/Zo Rl Date Copyright 2012
    Hearing Moral Reform: Abolitionists and Slave Music, 1838-1870 by Karl Byrn A thesis submitted to Sonoma State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS III History Dr. Steve Estes r Z- /Li ,/zO rL Date Copyright 2012 By Karl Bym ii AUTHORIZATION FOR REPRODUCTION OF MASTER'S THESIS/PROJECT I grant permission for the reproduction ofparts ofthis thesis without further authorization from me, on the condition that the person or agency requesting reproduction absorb the cost and provide proper acknowledgement of authorship. Permission to reproduce this thesis in its entirety must be obtained from me. iii Hearing Moral Reform: Abolitionists and Slave Music, 1838-1870 Thesis by Karl Byrn ABSTRACT Purpose ofthe Study: The purpose of this study is to analyze commentary on slave music written by certain abolitionists during the American antebellum and Civil War periods. The analysis is not intended to follow existing scholarship regarding the origins and meanings of African-American music, but rather to treat this music commentary as a new source for understanding the abolitionist movement. By treating abolitionists' reports of slave music as records of their own culture, this study will identify larger intellectual and cultural currents that determined their listening conditions and reactions. Procedure: The primary sources involved in this study included journals, diaries, letters, articles, and books written before, during, and after the American Civil War. I found, in these sources, that the responses by abolitionists to the music of slaves and freedmen revealed complex ideologies and motivations driven by religious and reform traditions, antebellum musical cultures, and contemporary racial and social thinking.
    [Show full text]
  • Lecture 10: Learning DNF, AC0, Juntas Feb 15, 2007 Lecturer: Ryan O’Donnell Scribe: Elaine Shi
    Analysis of Boolean Functions (CMU 18-859S, Spring 2007) Lecture 10: Learning DNF, AC0, Juntas Feb 15, 2007 Lecturer: Ryan O’Donnell Scribe: Elaine Shi 1 Learning DNF in Almost Polynomial Time From previous lectures, we have learned that if a function f is ǫ-concentrated on some collection , then we can learn the function using membership queries in poly( , 1/ǫ)poly(n) log(1/δ) time.S |S| O( w ) In the last lecture, we showed that a DNF of width w is ǫ-concentrated on a set of size n ǫ , and O( w ) concluded that width-w DNFs are learnable in time n ǫ . Today, we shall improve this bound, by showing that a DNF of width w is ǫ-concentrated on O(w log 1 ) a collection of size w ǫ . We shall hence conclude that poly(n)-size DNFs are learnable in almost polynomial time. Recall that in the last lecture we introduced H˚astad’s Switching Lemma, and we showed that 1 DNFs of width w are ǫ-concentrated on degrees up to O(w log ǫ ). Theorem 1.1 (Hastad’s˚ Switching Lemma) Let f be computable by a width-w DNF, If (I, X) is a random restriction with -probability ρ, then d N, ∗ ∀ ∈ d Pr[DT-depth(fX→I) >d] (5ρw) I,X ≤ Theorem 1.2 If f is a width-w DNF, then f(U)2 ǫ ≤ |U|≥OX(w log 1 ) ǫ b O(w log 1 ) To show that a DNF of width w is ǫ-concentrated on a collection of size w ǫ , we also need the following theorem: Theorem 1.3 If f is a width-w DNF, then 1 |U| f(U) 2 20w | | ≤ XU b Proof: Let (I, X) be a random restriction with -probability 1 .
    [Show full text]