arXiv subjects

Yuval Rabani

Publications and source records attributed to Yuval Rabani.

33 records · Page 2Linked to original sources

arXiv2020

Online Multiserver Convex Chasing and Optimization

We introduce the problem of $k$-chasing of convex functions, a simultaneous generalization of both the famous k-server problem in $R^d$, and of the problem of chasing convex bodies and functions. Aside from fundamental interest in this general form, it has natural applications to online $k$-clustering problems with objectives such as $k$-median or $k$-means. We show that this problem exhibits a rich landscape of behavior. In general, if both $k > 1$ and $d > 1$ there does not exist any online algorithm with bounded competitiveness. By contrast, we exhibit a class of nicely behaved functions (which include in particular the above-mentioned clustering problems), for which we show that competitive online algorithms exist, and moreover with dimension-free competitive ratio. We also introduce a parallel question of top-$k$ action regret minimization in the realm of online convex optimization. There, too, a much rougher landscape emerges for $k > 1$. While it is possible to achieve vanishing regret, unlike the top-one action case the rate of vanishing does not speed up for strongly convex functions. Moreover, vanishing regret necessitates both intractable computations and randomness. Finally we leave open whether almost dimension-free regret is achievable for $k > 1$ and general convex losses. As evidence that it might be possible, we prove dimension-free regret for linear losses via an information-theoretic argument.

cs.DS↗

arXiv2019

Parametrized Metrical Task Systems

We consider parametrized versions of metrical task systems and metrical service systems, two fundamental models of online computing, where the constrained parameter is the number of possible distinct requests $m$. Such parametrization occurs naturally in a wide range of applications. Striking examples are certain power management problems, which are modeled as metrical task systems with $m=2$. We characterize the competitive ratio in terms of the parameter $m$ for both deterministic and randomized algorithms on hierarchically separated trees. Our findings uncover a rich and unexpected picture that differs substantially from what is known or conjectured about the unparametrized versions of these problems. For metrical task systems, we show that deterministic algorithms do not exhibit any asymptotic gain beyond one-level trees (namely, uniform metric spaces), whereas randomized algorithms do not exhibit any asymptotic gain even for one-level trees. In contrast, the special case of metrical service systems (subset chasing) behaves very differently. Both deterministic and randomized algorithms exhibit gain, for $m$ sufficiently small compared to $n$, for any number of levels. Most significantly, they exhibit a large gain for uniform metric spaces and a smaller gain for two-level trees. Moreover, it turns out that in these cases (as well as in the case of metrical task systems for uniform metric spaces with $m$ being an absolute constant), deterministic algorithms are essentially as powerful as randomized algorithms. This is surprising and runs counter to the ubiquitous intuition/conjecture that, for most problems that can be modeled as metrical task systems, the randomized competitive ratio is polylogarithmic in the deterministic competitive ratio.

cs.DS↗

arXiv2017

Approximating Sparsest Cut in Low Rank Graphs via Embeddings from Approximately Low-Dimensional Spaces

We consider the problem of embedding a finite set of points $\{x_1, \ldots, x_n\} \in \mathbb{R}^d$ that satisfy $\ell_2^2$ triangle inequalities into $\ell_1$, when the points are approximately low-dimensional. Goemans (unpublished, appears in a work of [Magen and Moharammi, 2008]) showed that such points residing in \emph{exactly} $d$ dimensions can be embedded into $\ell_1$ with distortion at most $\sqrt{d}$. We prove the following robust analogue of this statement: if there exists a $r$-dimensional subspace $Π$ such that the projections onto this subspace satisfy $\sum_{i,j \in [n]}\Vert Πx_i - Πx_j \Vert _2^2 \geq Ω(1) \sum_{i,j \in [n]}\Vert x_i - x_j \Vert _2^2$, then there is an embedding of the points into $\ell_1$ with $O(\sqrt{r})$ average distortion. A consequence of this result is that the integrality gap of the well-known Goemans-Linial SDP relaxation for the Uniform Sparsest Cut problem is $O(\sqrt{r})$ on graphs $G$ whose $r$-th smallest normalized eigenvalue of the Laplacian satisfies $λ_r(G)/n \geq Ω(1)Φ_{SDP} (G)$. Our result improves upon the previously known bound of $O(r)$ on the average distortion, and the integrality gap of the Goemans-Linial SDP under the same preconditions, proven in the previous works of [Deshpande and Venkat, 2014] and [Deshpande, Harsha and Venkat, 2016].

cs.DS↗

arXiv2017

Strictly Balancing Matrices in Polynomial Time Using Osborne's Iteration

Osborne's iteration is a method for balancing $n\times n$ matrices which is widely used in linear algebra packages, as balancing preserves eigenvalues and stabilizes their numeral computation. The iteration can be implemented in any norm over $\mathbb{R}^n$, but it is normally used in the $L_2$ norm. The choice of norm not only affects the desired balance condition, but also defines the iterated balancing step itself. In this paper we focus on Osborne's iteration in any $L_p$ norm, where $p < \infty$. We design a specific implementation of Osborne's iteration in any $L_p$ norm that converges to a strictly $ε$-balanced matrix in $\tilde{O}(ε^{-2}n^{9} K)$ iterations, where $K$ measures, roughly, the {\em number of bits} required to represent the entries of the input matrix. This is the first result that proves that Osborne's iteration in the $L_2$ norm (or any $L_p$ norm, $p < \infty$) strictly balances matrices in polynomial time. This is a substantial improvement over our recent result (in SODA 2017) that showed weak balancing in $L_p$ norms. Previously, Schulman and Sinclair (STOC 2015) showed strong balancing of Osborne's iteration in the $L_\infty$ norm. Their result does not imply any bounds on strict balancing in other norms.

cs.DS↗

arXiv2016

Matrix Balancing in Lp Norms: A New Analysis of Osborne's Iteration

We study an iterative matrix conditioning algorithm due to Osborne (1960). The goal of the algorithm is to convert a square matrix into a balanced matrix where every row and corresponding column have the same norm. The original algorithm was proposed for balancing rows and columns in the $L_2$ norm, and it works by iterating over balancing a row-column pair in fixed round-robin order. Variants of the algorithm for other norms have been heavily studied and are implemented as standard preconditioners in many numerical linear algebra packages. Recently, Schulman and Sinclair (2015), in a first result of its kind for any norm, analyzed the rate of convergence of a variant of Osborne's algorithm that uses the $L_{\infty}$ norm and a different order of choosing row-column pairs. In this paper we study matrix balancing in the $L_1$ norm and other $L_p$ norms. We show the following results for any matrix $A = (a_{ij})_{i,j=1}^n$, resolving in particular a main open problem mentioned by Schulman and Sinclair. 1) We analyze the iteration for the $L_1$ norm under a greedy order of balancing. We show that it converges to an $ε$-balanced matrix in $K = O(\min\{ε^{-2}\log w,ε^{-1}n^{3/2}\log(w/ε)\})$ iterations that cost a total of $O(m + Kn\log n)$ arithmetic operations over $O(n\log w)$-bit numbers. Here $m$ is the number of non-zero entries of $A$, and $w = \sum_{i,j} |a_{ij}|/a_{\min}$ with $a_{\min} = \min\{|a_{ij}|:\ a_{ij}\neq 0\}$. 2) We show that the original round-robin implementation converges to an $ε$-balanced matrix in $O(ε^{-2}n^2\log w)$ iterations totalling $O(ε^{-2}mn\log w)$ arithmetic operations over $O(n\log w)$-bit numbers. 3) We demonstrate a lower bound of $Ω(1/\sqrtε)$ on the convergence rate of any implementation of the iteration.

cs.DS↗

arXiv2016

Market Dynamics of Best-Response with Lookahead

One attractive approach to market dynamics is the level $k$ model in which a level $0$ player adopts a very simple response to current conditions, a level $1$ player best-responds to a model in which others take level $0$ actions, and so forth. (This is analogous to $k$-ply exploration of game trees in AI, and to receding-horizon control in control theory.) If players have deterministic mental models with this kind of finite-level response, there is obviously no way their mental models can all be consistent. Nevertheless, there is experimental evidence that people act this way in many situations, motivating the question of what the dynamics of such interactions lead to. We address this question in the setting of Fisher Markets with constant elasticities of substitution (CES) utilities, in the weak gross substitutes (WGS) regime. We show that despite the inconsistency of the mental models, and even if players' models change arbitrarily from round to round, the market converges to its unique equilibrium. (We show this for both synchronous and asynchronous discrete-time updates.) Moreover, the result is computationally feasible in the sense that the convergence rate is linear, i.e., the distance to equilibrium decays exponentially fast. To the best of our knowledge, this is the first result that demonstrates, in Fisher markets, convergence at any rate for dynamics driven by a plausible model of seller incentives. Even for the simple case of (level $0$) best-response dynamics, where we observe that convergence at some rate can be derived from recent results in convex optimization, our result is the first to demonstrate a linear rate of convergence.

cs.GT↗

arXiv2016

The Invisible Hand of Laplace: the Role of Market Structure in Price Convergence and Oscillation

A fundamental question about a market is under what conditions, and then how rapidly, does price signaling cause price equilibration. Qualitatively, this ought to depend on how well-connected the market is. We address this question quantitatively for a certain class of Arrow-Debreu markets with continuous-time proportional tâtonnement dynamics. We show that the algebraic connectivity of the market determines the effectiveness of price signaling equilibration. This also lets us study the rate of external noise that a market can tolerate and still maintain near-equilibrium prices.

econ.GN↗

arXiv2015

On Lipschitz extension from finite subsets

We prove that for every $n\in \mathbb{N}$ there exists a metric space $(X,d_X)$, an $n$-point subset $S\subseteq X$, a Banach space $(Z,\|\cdot\|_Z)$ and a $1$-Lipschitz function $f:S\to Z$ such that the Lipschitz constant of every function $F:X\to Z$ that extends $f$ is at least a constant multiple of $\sqrt{\log n}$. This improves a bound of Johnson and Lindenstrauss. We also obtain the following quantitative counterpart to a classical extension theorem of Minty. For every $α\in (1/2,1]$ and $n\in \mathbb{N}$ there exists a metric space $(X,d_X)$, an $n$-point subset $S\subseteq X$ and a function $f:S\to \ell_2$ that is $α$-Hölder with constant $1$, yet the $α$-Hölder constant of any $F:X\to \ell_2$ that extends $f$ satisfies $$ \|F\|_{\mathrm{Lip}(α)}\gtrsim (\log n)^{\frac{2α-1}{4α}}+\left(\frac{\log n}{\log\log n}\right)^{α^2-\frac12}. $$ We formulate a conjecture whose positive solution would strengthen Ball's nonlinear Maurey extension theorem, serving as a far-reaching nonlinear version of a theorem of König, Retherford and Tomczak-Jaegermann. We explain how this conjecture would imply as special cases answers to longstanding open questions of Johnson and Lindenstrauss and Kalton.

math.MG↗

arXiv2015

Convergence of Tâtonnement in Fisher Markets

Analyzing simple and natural price-adjustment processes that converge to a market equilibrium is a fundamental question in economics. Such an analysis may have implications in economic theory, computational economics, and distributed systems. Tâtonnement, proposed by Walras in 1874, is a process by which prices go up in response to excess demand, and down in response to excess supply. This paper analyzes the convergence of a time-discrete tâtonnement process, a problem that recently attracted considerable attention of computer scientists. We prove that the simple tâtonnement process that we consider converges (efficiently) to equilibrium prices and allocation in markets with nested CES-Leontief utilities, generalizing some of the previous convergence proofs for more restricted types of utility functions.

cs.GT↗

arXiv2015

Learning Arbitrary Statistical Mixtures of Discrete Distributions

We study the problem of learning from unlabeled samples very general statistical mixture models on large finite sets. Specifically, the model to be learned, $\vartheta$, is a probability distribution over probability distributions $p$, where each such $p$ is a probability distribution over $[n] = \{1,2,\dots,n\}$. When we sample from $\vartheta$, we do not observe $p$ directly, but only indirectly and in very noisy fashion, by sampling from $[n]$ repeatedly, independently $K$ times from the distribution $p$. The problem is to infer $\vartheta$ to high accuracy in transportation (earthmover) distance. We give the first efficient algorithms for learning this mixture model without making any restricting assumptions on the structure of the distribution $\vartheta$. We bound the quality of the solution as a function of the size of the samples $K$ and the number of samples used. Our model and results have applications to a variety of unsupervised learning scenarios, including learning topic models and collaborative filtering.

cs.LG↗

arXiv2013

Learning Mixtures of Arbitrary Distributions over Large Discrete Domains

We give an algorithm for learning a mixture of {\em unstructured} distributions. This problem arises in various unsupervised learning scenarios, for example in learning {\em topic models} from a corpus of documents spanning several topics. We show how to learn the constituents of a mixture of $k$ arbitrary distributions over a large discrete domain $[n]=\{1,2,\dots,n\}$ and the mixture weights, using $O(n\polylog n)$ samples. (In the topic-model learning setting, the mixture constituents correspond to the topic distributions.) This task is information-theoretically impossible for $k>1$ under the usual sampling process from a mixture distribution. However, there are situations (such as the above-mentioned topic model case) in which each sample point consists of several observations from the same mixture constituent. This number of observations, which we call the {\em "sampling aperture"}, is a crucial parameter of the problem. We obtain the {\em first} bounds for this mixture-learning problem {\em without imposing any assumptions on the mixture constituents.} We show that efficient learning is possible exactly at the information-theoretically least-possible aperture of $2k-1$. Thus, we achieve near-optimal dependence on $n$ and optimal aperture. While the sample-size required by our algorithm depends exponentially on $k$, we prove that such a dependence is {\em unavoidable} when one considers general mixtures. A sequence of tools contribute to the algorithm, such as concentration results for random matrices, dimension reduction, moment estimations, and sensitivity analysis.

cs.LG↗

arXiv2013

An Optimal Randomized Online Algorithm for Reordering Buffer Management

We give an $O(\log\log k)$-competitive randomized online algorithm for reordering buffer management, where $k$ is the buffer size. Our bound matches the lower bound of Adamaszek et al. (STOC 2011). Our algorithm has two stages which are executed online in parallel. The first stage computes deterministically a feasible fractional solution to an LP relaxation for reordering buffer management. The second stage "rounds" using randomness the fractional solution. The first stage is based on the online primal-dual schema, combined with a dual fitting argument. As multiplicative weights steps and dual fitting steps are interleaved and in some sense conflicting, combining them is challenging. We also note that we apply the primal-dual schema to a relaxation with mixed packing and covering constraints. We pay the $O(\log\log k)$ competitive factor for the gap between the computed LP solution and the optimal LP solution. The second stage gives an online algorithm that converts the LP solution to an integral solution, while increasing the cost by an O(1) factor. This stage generalizes recent results that gave a similar approximation factor for rounding the LP solution, albeit using an offline rounding algorithm.

cs.DS↗

arXiv2012

A Constant Factor Approximation Algorithm for Reordering Buffer Management

In the reordering buffer management problem (RBM) a sequence of $n$ colored items enters a buffer with limited capacity $k$. When the buffer is full, one item is removed to the output sequence, making room for the next input item. This step is repeated until the input sequence is exhausted and the buffer is empty. The objective is to find a sequence of removals that minimizes the total number of color changes in the output sequence. The problem formalizes numerous applications in computer and production systems, and is known to be NP-hard. We give the first constant factor approximation guarantee for RBM. Our algorithm is based on an intricate "rounding" of the solution to an LP relaxation for RBM, so it also establishes a constant upper bound on the integrality gap of this relaxation. Our results improve upon the best previous bound of $O(\sqrt{\log k})$ of Adamaszek et al. (STOC 2011) that used different methods and gave an online algorithm. Our constant factor approximation beats the super-constant lower bounds on the competitive ratio given by Adamaszek et al. This is the first demonstration of an offline algorithm for RBM that is provably better than any online algorithm.

cs.DS↗

arXiv2011

On Parsimonious Explanations for 2-D Tree- and Linearly-Ordered Data

This paper studies the "explanation problem" for tree- and linearly-ordered array data, a problem motivated by database applications and recently solved for the one-dimensional tree-ordered case. In this paper, one is given a matrix A whose rows and columns have semantics: special subsets of the rows and special subsets of the columns are meaningful, others are not. A submatrix in A is said to be meaningful if and only if it is the cross product of a meaningful row subset and a meaningful column subset, in which case we call it an "allowed rectangle." The goal is to "explain" A as a sparse sum of weighted allowed rectangles. Specifically, we wish to find as few weighted allowed rectangles as possible such that, for all i,j, a_{ij} equals the sum of the weights of all rectangles which include cell (i,j). In this paper we consider the natural cases in which the matrix dimensions are tree-ordered or linearly-ordered. In the tree-ordered case, we are given a rooted tree T1 whose leaves are the rows of A and another, T2, whose leaves are the columns. Nodes of the trees correspond in an obvious way to the sets of their leaf descendants. In the linearly-ordered case, a set of rows or columns is meaningful if and only if it is contiguous. For tree-ordered data, we prove the explanation problem NP-Hard and give a randomized 2-approximation algorithm for it. For linearly-ordered data, we prove the explanation problem NP-Hard and give a 2.56-approximation algorithm. To our knowledge, these are the first results for the problem of sparsely and exactly representing matrices by weighted rectangles.

cs.DS↗

arXiv2010

Rademacher Chaos, Random Eulerian Graphs and The Sparse Johnson-Lindenstrauss Transform

The celebrated dimension reduction lemma of Johnson and Lindenstrauss has numerous computational and other applications. Due to its application in practice, speeding up the computation of a Johnson-Lindenstrauss style dimension reduction is an important question. Recently, Dasgupta, Kumar, and Sarlos (STOC 2010) constructed such a transform that uses a sparse matrix. This is motivated by the desire to speed up the computation when applied to sparse input vectors, a scenario that comes up in applications. The sparsity of their construction was further improved by Kane and Nelson (ArXiv 2010). We improve the previous bound on the number of non-zero entries per column of Kane and Nelson from $O(1/ε\log(1/δ)\log(k/δ))$ (where the target dimension is $k$, the distortion is $1\pm ε$, and the failure probability is $δ$) to $$ O\left({1\overε} \left({\log(1/δ)\log\log\log(1/δ) \over \log\log(1/δ)}\right)^2\right). $$ We also improve the amount of randomness needed to generate the matrix. Our results are obtained by connecting the moments of an order 2 Rademacher chaos to the combinatorial properties of random Eulerian multigraphs. Estimating the chance that a random multigraph is composed of a given number of node-disjoint Eulerian components leads to a new tail bound on the chaos. Our estimates may be of independent interest, and as this part of the argument is decoupled from the analysis of the coefficients of the chaos, we believe that our methods can be useful in the analysis of other chaoses.

cs.DS↗