arXiv subjects

Venkatesan Guruswami

Publications and source records attributed to Venkatesan Guruswami.

At least 127 records · Page 7Linked to original sources

arXiv2018

An Algorithmic Blend of LPs and Ring Equations for Promise CSPs

Promise CSPs are a relaxation of constraint satisfaction problems where the goal is to find an assignment satisfying a relaxed version of the constraints. Several well-known problems can be cast as promise CSPs including approximate graph coloring, discrepancy minimization, and interesting variants of satisfiability. Similar to CSPs, the tractability of promise CSPs can be tied to the structure of operations on the solution space called polymorphisms, though in the promise world these operations are much less constrained. Under the thesis that non-trivial polymorphisms govern tractability, promise CSPs therefore provide a fertile ground for the discovery of novel algorithms. In previous work, we classified Boolean promise CSPs when the constraint predicates are symmetric. In this work, we vastly generalize these algorithmic results. Specifically, we show that promise CSPs that admit a family of "regional-periodic" polymorphisms are in P, assuming that determining which region a point is in can be computed in polynomial time. Such polymorphisms are quite general and are obtained by gluing together several functions that are periodic in the Hamming weights in different blocks of the input. Our algorithm is based on a novel combination of linear programming and solving linear systems over rings. We also abstract a framework based on reducing a promise CSP to a CSP over an infinite domain, solving it there, and then rounding the solution to an assignment for the promise CSP instance. The rounding step is intimately tied to the family of polymorphisms and clarifies the connection between polymorphisms and algorithms in this context. As a key ingredient, we introduce the technique of finding a solution to a linear program with integer coefficients that lies in a different ring (such as $\mathbb Z[\sqrt{2}]$) to bypass ad-hoc adjustments for lying on a rounding boundary.

cs.CC↗

arXiv2018

How long can optimal locally repairable codes be?

A locally repairable code (LRC) with locality $r$ allows for the recovery of any erased codeword symbol using only $r$ other codeword symbols. A Singleton-type bound dictates the best possible trade-off between the dimension and distance of LRCs --- an LRC attaining this trade-off is deemed \emph{optimal}. Such optimal LRCs have been constructed over alphabets growing linearly in the block length. Unlike the classical Singleton bound, however, it was not known if such a linear growth in the alphabet size is necessary, or for that matter even if the alphabet needs to grow at all with the block length. Indeed, for small code distances $3,4$, arbitrarily long optimal LRCs were known over fixed alphabets. Here, we prove that for distances $d \ge 5$, the code length $n$ of an optimal LRC over an alphabet of size $q$ must be at most roughly $O(d q^3)$. For the case $d=5$, our upper bound is $O(q^2)$. We complement these bounds by showing the existence of optimal LRCs of length $Ω_{d,r}(q^{1+1/\lfloor(d-3)/2\rfloor})$ when $d \le r+2$. These bounds match when $d=5$, thus pinning down $n=Θ(q^2)$ as the asymptotically largest length of an optimal LRC for this case.

cs.IT↗

arXiv2018

$ε$-MSR Codes: Contacting Fewer Code Blocks for Exact Repair

$ε$-Minimum Storage Regenerating ($ε$-MSR) codes form a special class of Maximum Distance Separable (MDS) codes, providing mechanisms for exact regeneration of a single code block in their codewords by downloading slighly sub-optimal amount of information from the remaining code blocks. The key advantage of these codes is a significantly lower sub-packetization that grows only logarithmically with the length of the code, while providing optimality in storage and error-correcting capacity. However, from an implementation point of view, these codes require each remaining code block to be available for the repair of any single code block. In this paper, we address this issue by constructing $ε$-MSR codes that can repair a failed code block by contacting a fewer number of available code blocks. When a code block fails, our repair procedure needs to contact a few compulsory code blocks and is free to choose any subset of available code blocks for the remaining choices. Further, our construction requiresa field size linear in code length and ensures load balancing among the contacted code blocks in terms of information downloaded from them for a single repair.

cs.IT↗

arXiv2018

Beating Fredman-Komlós for perfect $k$-hashing

We say a subset $C \subseteq \{1,2,\dots,k\}^n$ is a $k$-hash code (also called $k$-separated) if for every subset of $k$ codewords from $C$, there exists a coordinate where all these codewords have distinct values. Understanding the largest possible rate (in bits), defined as $(\log_2 |C|)/n$, of a $k$-hash code is a classical problem. It arises in two equivalent contexts: (i) the smallest size possible for a perfect hash family that maps a universe of $N$ elements into $\{1,2,\dots,k\}$, and (ii) the zero-error capacity for decoding with lists of size less than $k$ for a certain combinatorial channel. A general upper bound of $k!/k^{k-1}$ on the rate of a $k$-hash code (in the limit of large $n$) was obtained by Fredman and Komlós in 1984 for any $k \geq 4$. While better bounds have been obtained for $k=4$, their original bound has remained the best known for each $k \ge 5$. In this work, we obtain the first improvement to the Fredman-Komlós bound for every $k \ge 5$. While we get explicit (numerical) bounds for $k=5,6$, for larger $k$ we only show that the FK bound can be improved by a positive, but unspecified, amount. Under a conjecture on the optimum value of a certain polynomial optimization problem over the simplex, our methods allow an effective bound to be computed for every $k$.

cs.IT↗

arXiv2017

On the List-Decodability of Random Linear Rank-Metric Codes

The list-decodability of random linear rank-metric codes is shown to match that of random rank-metric codes. Specifically, an $\mathbb{F}_q$-linear rank-metric code over $\mathbb{F}_q^{m \times n}$ of rate $R = (1-ρ)(1-\frac{n}{m}ρ)-\varepsilon$ is shown to be (with high probability) list-decodable up to fractional radius $ρ\in (0,1)$ with lists of size at most $\frac{C_{ρ,q}}{\varepsilon}$, where $C_{ρ,q}$ is a constant depending only on $ρ$ and $q$. This matches the bound for random rank-metric codes (up to constant factors). The proof adapts the approach of Guruswami, Håstad, Kopparty (STOC 2010), who established a similar result for the Hamming metric case, to the rank-metric setting.

cs.CC↗

arXiv2017

MDS Code Constructions with Small Sub-packetization and Near-optimal Repair Bandwidth

This paper addresses the problem of constructing MDS codes that enable exact repair of each code block with small repair bandwidth, which refers to the total amount of information flow from the remaining code blocks during the repair process. This problem naturally arises in the context of distributed storage systems as the node repair problem [7]. The constructions of exact-repairable MDS codes with optimal repair-bandwidth require working with large sub-packetization levels, which restricts their employment in practice. This paper presents constructions for MDS codes that simultaneously provide both small repair bandwidth and small sub-packetization level. In particular, this paper presents two general approaches to construct exact-repairable MDS codes that aim at significantly reducing the required sub-packetization level at the cost of slightly sub-optimal repair bandwidth. The first approach gives MDS codes that have repair bandwidth at most twice the optimal repair-bandwidth. Additionally, these codes also have the smallest possible sub-packetization level $\ell = O(r)$, where $r$ denotes the number of parity blocks. This approach is then generalized to design codes that have their repair bandwidth approaching the optimal repair-bandwidth at the cost of graceful increment in the required sub-packetization level. The second approach transforms an MDS code with optimal repair-bandwidth and large sub-packetization level into a longer MDS code with small sub-packetization level and near-optimal repair bandwidth. For a given $r$, the obtained codes have their sub-packetization level scaling logarithmically with the code length. In addition, the obtained codes require field size only linear in the code length and ensure load balancing among the intact code blocks in terms of the information downloaded from these blocks during the exact reconstruction of a code block.

cs.IT↗

arXiv2017

Coding against deletions in oblivious and online models

We consider binary error correcting codes when errors are deletions. A basic challenge concerning deletion codes is determining $p_0^{(adv)}$, the zero-rate threshold of adversarial deletions, defined to be the supremum of all $p$ for which there exists a code family with rate bounded away from 0 capable of correcting a fraction $p$ of adversarial deletions. A recent construction of deletion-correcting codes [Bukh et al 17] shows that $p_0^{(adv)} \ge \sqrt{2}-1$, and the trivial upper bound, $p_0^{(adv)}\le\frac{1}{2}$, is the best known. Perhaps surprisingly, we do not know whether or not $p_0^{(adv)} = 1/2$. In this work, to gain further insight into deletion codes, we explore two related error models: oblivious deletions and online deletions, which are in between random and adversarial deletions in power. In the oblivious model, the channel can inflict an arbitrary pattern of $pn$ deletions, picked without knowledge of the codeword. We prove the existence of binary codes of positive rate that can correct any fraction $p < 1$ of oblivious deletions, establishing that the associated zero-rate threshold $p_0^{(obliv)}$ equals $1$. For online deletions, where the channel decides whether to delete bit $x_i$ based only on knowledge of bits $x_1x_2\dots x_i$, define the deterministic zero-rate threshold for online deletions $p_0^{(on,d)}$ to be the supremum of $p$ for which there exist deterministic codes against an online channel causing $pn$ deletions with low average probability of error. That is, the probability that a randomly chosen codeword is decoded incorrectly is small. We prove $p_0^{(adv)}=\frac{1}{2}$ if and only if $p_0^{(on,d)}=\frac{1}{2}$.

cs.IT↗

arXiv2017

Sum-of-Squares Certificates for Maxima of Random Tensors on the Sphere

For an $n$-variate order-$d$ tensor $A$, define $ A_{\max} := \sup_{\| x \|_2 = 1} \langle A , x^{\otimes d} \rangle$ to be the maximum value taken by the tensor on the unit sphere. It is known that for a random tensor with i.i.d $\pm 1$ entries, $A_{\max} \lesssim \sqrt{n\cdot d\cdot\log d}$ w.h.p. We study the problem of efficiently certifying upper bounds on $A_{\max}$ via the natural relaxation from the Sum of Squares (SoS) hierarchy. Our results include: - When $A$ is a random order-$q$ tensor, we prove that $q$ levels of SoS certifies an upper bound $B$ on $A_{\max}$ that satisfies \[ B ~~~~\leq~~ A_{\max} \cdot \biggl(\frac{n}{q^{\,1-o(1)}}\biggr)^{q/4-1/2} \quad \text{w.h.p.} \] Our upper bound improves a result of Montanari and Richard (NIPS 2014) when $q$ is large. - We show the above bound is the best possible up to lower order terms, namely the optimum of the level-$q$ SoS relaxation is at least \[ A_{\max} \cdot \biggl(\frac{n}{q^{\,1+o(1)}}\biggr)^{q/4-1/2} \ . \] - When $A$ is a random order-$d$ tensor, we prove that $q$ levels of SoS certifies an upper bound $B$ on $A_{\max}$ that satisfies \[ B ~~\leq ~~ A_{\max} \cdot \biggl(\frac{\widetilde{O}(n)}{q}\biggr)^{d/4 - 1/2} \quad \text{w.h.p.} \] For growing $q$, this improves upon the bound certified by constant levels of SoS. This answers in part, a question posed by Hopkins, Shi, and Steurer (COLT 2015), who established the tight characterization for constant levels of SoS.

cs.CC↗

arXiv2017

Locality via Partially Lifted Codes

In error-correcting codes, locality refers to several different ways of quantifying how easily a small amount of information can be recovered from encoded data. In this work, we study a notion of locality called the s-Disjoint-Repair-Group Property (s-DRGP). This notion can interpolate between two very different settings in coding theory: that of Locally Correctable Codes (LCCs) when s is large---a very strong guarantee---and Locally Recoverable Codes (LRCs) when s is small---a relatively weaker guarantee. This motivates the study of the s-DRGP for intermediate s, which is the focus of our paper. We construct codes in this parameter regime which have a higher rate than previously known codes. Our construction is based on a novel variant of the lifted codes of Guo, Kopparty and Sudan. Beyond the results on the s-DRGP, we hope that our construction is of independent interest, and will find uses elsewhere.

cs.IT↗

arXiv2017

Weak Decoupling, Polynomial Folds, and Approximate Optimization over the Sphere

We consider the following basic problem: given an $n$-variate degree-$d$ homogeneous polynomial $f$ with real coefficients, compute a unit vector $x \in \mathbb{R}^n$ that maximizes $|f(x)|$. Besides its fundamental nature, this problem arises in diverse contexts ranging from tensor and operator norms to graph expansion to quantum information theory. The homogeneous degree $2$ case is efficiently solvable as it corresponds to computing the spectral norm of an associated matrix, but the higher degree case is NP-hard. We give approximation algorithms for this problem that offer a trade-off between the approximation ratio and running time: in $n^{O(q)}$ time, we get an approximation within factor $O_d((n/q)^{d/2-1})$ for arbitrary polynomials, $O_d((n/q)^{d/4-1/2})$ for polynomials with non-negative coefficients, and $O_d(\sqrt{m/q})$ for sparse polynomials with $m$ monomials. The approximation guarantees are with respect to the optimum of the level-$q$ sum-of-squares (SoS) SDP relaxation of the problem. Known polynomial time algorithms for this problem rely on "decoupling lemmas." Such tools are not capable of offering a trade-off like our results as they blow up the number of variables by a factor equal to the degree. We develop new decoupling tools that are more efficient in the number of variables at the expense of less structure in the output polynomials. This enables us to harness the benefits of higher level SoS relaxations. We complement our algorithmic results with some polynomially large integrality gaps, albeit for a slightly weaker (but still very natural) relaxation. Toward this, we give a method to lift a level-$4$ solution matrix $M$ to a higher level solution, under a mild technical condition on $M$.

cs.DS↗

arXiv2017

Subspace Designs based on Algebraic Function Fields

Subspace designs are a (large) collection of high-dimensional subspaces $\{H_i\}$ of $\F_q^m$ such that for any low-dimensional subspace $W$, only a small number of subspaces from the collection have non-trivial intersection with $W$; more precisely, the sum of dimensions of $W \cap H_i$ is at most some parameter $L$. The notion was put forth by Guruswami and Xing (STOC'13) with applications to list decoding variants of Reed-Solomon and algebraic-geometric codes, and later also used for explicit rank-metric codes with optimal list decoding radius. Guruswami and Kopparty (FOCS'13, Combinatorica'16) gave an explicit construction of subspace designs with near-optimal parameters. This construction was based on polynomials and has close connections to folded Reed-Solomon codes, and required large field size (specifically $q \ge m$). Forbes and Guruswami (RANDOM'15) used this construction to give explicit constant degree "dimension expanders" over large fields, and noted that subspace designs are a powerful tool in linear-algebraic pseudorandomness. Here, we construct subspace designs over any field, at the expense of a modest worsening of the bound $L$ on total intersection dimension. Our approach is based on a (non-trivial) extension of the polynomial-based construction to algebraic function fields, and instantiating the approach with cyclotomic function fields. Plugging in our new subspace designs in the construction of Forbes and Guruswami yields dimension expanders over $\F^n$ for any field $\F$, with logarithmic degree and expansion guarantee for subspaces of dimension $Ω(n/(\log \log n))$.

cs.CC↗

arXiv2017

Efficiently list-decodable punctured Reed-Muller codes

The Reed-Muller (RM) code encoding $n$-variate degree-$d$ polynomials over ${\mathbb F}_q$ for $d < q$, with its evaluation on ${\mathbb F}_q^n$, has relative distance $1-d/q$ and can be list decoded from a $1-O(\sqrt{d/q})$ fraction of errors. In this work, for $d \ll q$, we give a length-efficient puncturing of such codes which (almost) retains the distance and list decodability properties of the Reed-Muller code, but has much better rate. Specificially, when $q =Ω( d^2/ε^2)$, we given an explicit rate $Ω\left(\fracε{d!}\right)$ puncturing of Reed-Muller codes which have relative distance at least $(1-ε)$ and efficient list decoding up to $(1-\sqrtε)$ error fraction. This almost matches the performance of random puncturings which work with the weaker field size requirement $q= Ω( d/ε^2)$. We can also improve the field size requirement to the optimal (up to constant factors) $q =Ω( d/ε)$, at the expense of a worse list decoding radius of $1-ε^{1/3}$ and rate $Ω\left(\frac{ε^2}{d!}\right)$. The first of the above trade-offs is obtained by substituting for the variables functions with carefully chosen pole orders from an algebraic function field; this leads to a puncturing for which the RM code is a subcode of a certain algebraic-geometric code (which is known to be efficiently list decodable). The second trade-off is obtained by concatenating this construction with a Reed-Solomon based multiplication friendly pair, and using the list recovery property of algebraic-geometric codes.

cs.IT↗

arXiv2016

Repairing Reed-Solomon Codes

We study the performance of Reed-Solomon (RS) codes for the \em exact repair problem \em in distributed storage. Our main result is that, in some parameter regimes, Reed-Solomon codes are optimal regenerating codes, among MDS codes with linear repair schemes. Moreover, we give a characterization of MDS codes with linear repair schemes which holds in any parameter regime, and which can be used to give non-trivial repair schemes for RS codes in other settings. More precisely, we show that for $k$-dimensional RS codes whose evaluation points are a finite field of size $n$, there are exact repair schemes with bandwidth $(n-1)\log((n-1)/(n-k))$ bits, and that this is optimal for any MDS code with a linear repair scheme. In contrast, the naive (commonly implemented) repair algorithm for this RS code has bandwidth $k\log(n)$ bits. When the entire field is used as evaluation points, the number of nodes $n$ is much larger than the number of bits per node (which is $O(\log(n))$), and so this result holds only when the degree of sub-packetization is small. However, our method applies in any parameter regime, and to illustrate this for high levels of sub-packetization we give an improved repair scheme for a specific (14,10)-RS code used in the Facebook Hadoop Analytics cluster.

cs.IT↗

arXiv2016

New MDS codes with small sub-packetization and near-optimal repair bandwidth

An $(n, M)$ vector code $\mathcal{C} \subseteq \mathbb{F}^n$ is a collection of $M$ codewords where $n$ elements (from the field $\mathbb{F}$) in each of the codewords are referred to as code blocks. Assuming that $\mathbb{F} \cong \mathbb{B}^{\ell}$, the code blocks are treated as $\ell$-length vectors over the base field $\mathbb{B}$. Equivalently, the code is said to have the sub-packetization level $\ell$. This paper addresses the problem of constructing MDS vector codes which enable exact reconstruction of each code block by downloading small amount of information from the remaining code blocks. The repair bandwidth of a code measures the information flow from the remaining code blocks during the reconstruction of a single code block. This problem naturally arises in the context of distributed storage systems as the node repair problem [4]. Assuming that $M = |\mathbb{B}|^{k\ell}$, the repair bandwidth of an MDS vector code is lower bounded by $\big(\frac{n - 1}{n - k}\big)\cdot \ell$ symbols (over the base field $\mathbb{B}$) which is also referred to as the cut-set bound [4]. For all values of $n$ and $k$, the MDS vector codes that attain the cut-set bound with the sub-packetization level $\ell = (n-k)^{\lceil{{n}/{(n-k)}}\rceil}$ are known in the literature [23, 35]. This paper presents a construction for MDS vector codes which simultaneously ensures both small repair bandwidth and small sub-packetization level. The obtained codes have the smallest possible sub-packetization level $\ell = O(n - k)$ for an MDS vector code and the repair bandwidth which is at most twice the cut-set bound. The paper then generalizes this code construction so that the repair bandwidth of the obtained codes approach the cut-set bound at the cost of increased sub-packetization level. The constructions presented in this paper give MDS vector codes which are linear over the base field $\mathbb{B}$.

cs.IT↗

arXiv2016

Efficiently decodable insertion/deletion codes for high-noise and high-rate regimes

This work constructs codes that are efficiently decodable from a constant fraction of \emph{worst-case} insertion and deletion errors in three parameter settings: (i) Binary codes with rate approaching 1; (ii) Codes with constant rate for error fraction approaching 1 over fixed alphabet size; and (iii) Constant rate codes over an alphabet of size $k$ for error fraction approaching $(k-1)/(k+1)$. When errors are constrained to deletions alone, efficiently decodable codes in each of these regimes were constructed recently. We complete the picture by constructing similar codes that are efficiently decodable in the insertion/deletion regime.

cs.IT↗

arXiv2016

Rapidly Mixing Markov Chains: A Comparison of Techniques (A Survey)

We survey existing techniques to bound the mixing time of Markov chains. The mixing time is related to a geometric parameter called conductance which is a measure of edge-expansion. Bounds on conductance are typically obtained by a technique called "canonical paths" where the idea is to find a set of paths, one between every source-destination pair, such that no edge is heavily congested. However, the canonical paths approach cannot always show rapid mixing of a rapidly mixing chain. This drawback disappears if we allow the flow between a pair of states to be spread along multiple paths. We prove that for a large class of Markov chains canonical paths does capture rapid mixing. Allowing multiple paths to route the flow still does help a great deal in proofs, as illustrated by a result of Morris & Sinclair (FOCS'99) on the rapid mixing of a Markov chain for sampling 0/1 knapsack solutions. A different approach to prove rapid mixing is "Coupling". Path Coupling is a variant discovered by Bubley & Dyer (FOCS'97) that often tremendously reduces the complexity of designing good Couplings. We present several applications of Path Coupling in proofs of rapid mixing. These invariably lead to much better bounds on mixing time than known using conductance, and moreover Coupling based proofs are typically simpler. This motivates the question of whether Coupling can be made to work whenever the chain is rapidly mixing. This question was answered in the negative by Kumar & Ramesh (FOCS'99), who showed that no Coupling strategy can prove the rapid mixing of the Jerrum-Sinclair chain for sampling perfect and near-perfect matchings.

cs.DS↗

arXiv2016

An improved bound on the fraction of correctable deletions

We consider codes over fixed alphabets against worst-case symbol deletions. For any fixed $k \ge 2$, we construct a family of codes over alphabet of size $k$ with positive rate, which allow efficient recovery from a worst-case deletion fraction approaching $1-\frac{2}{k+\sqrt k}$. In particular, for binary codes, we are able to recover a fraction of deletions approaching $1/(\sqrt 2 +1)=\sqrt 2-1 \approx 0.414$. Previously, even non-constructively the largest deletion fraction known to be correctable with positive rate was $1-Θ(1/\sqrt{k})$, and around $0.17$ for the binary case. Our result pins down the largest fraction of correctable deletions for $k$-ary codes as $1-Θ(1/k)$, since $1-1/k$ is an upper bound even for the simpler model of erasures where the locations of the missing symbols are known. Closing the gap between $(\sqrt 2 -1)$ and $1/2$ for the limit of worst-case deletions correctable by binary codes remains a tantalizing open question.

cs.IT↗

arXiv2016

Superlinear lower bounds for multipass graph processing

We prove $n^{1+Ω(1/p)}/p^{O(1)}$ lower bounds for the space complexity of $p$-pass streaming algorithms solving the following problems on $n$-vertex graphs: * testing if an undirected graph has a perfect matching (this implies lower bounds for computing a maximum matching or even just the maximum matching size), * testing if two specific vertices are at distance at most $2(p+1)$ in an undirected graph, * testing if there is a directed path from $s$ to $t$ for two specific vertices $s$ and $t$ in a directed graph. Prior to our result, it was known that these problems require $Ω(n^2)$ space in one pass, but no $n^{1+Ω(1)}$ lower bound was known for any $p\ge 2$. These streaming results follow from a communication complexity lower bound for a communication game in which the players hold two graphs on the same set of vertices. The task of the players is to find out whether the sets of vertices at distance exactly $p+1$ from a specific vertex intersect. The game requires a significant amount of communication only if the players are forced to speak in a specific difficult order. This is reminiscent of lower bounds for communication problems such as indexing and pointer chasing. Among other things, our line of attack requires proving an information cost lower bound for a decision version of the classic pointer chasing problem and a direct sum type theorem for the disjunction of several instances of this problem.

cs.CC↗