arXiv subjects

Roy Schwartz

Publications and source records attributed to Roy Schwartz.

86 records · Page 5Linked to original sources

arXiv2018

A Tight Approximation for Submodular Maximization with Mixed Packing and Covering Constraints

Motivated by applications in machine learning, such as subset selection and data summarization, we consider the problem of maximizing a monotone submodular function subject to mixed packing and covering constraints. We present a tight approximation algorithm that for any constant $ε>0$ achieves a guarantee of $1-\frac{1}{\mathrm{e}}-ε$ while violating only the covering constraints by a multiplicative factor of $1-ε$. Our algorithm is based on a novel enumeration method, which unlike previous known enumeration techniques, can handle both packing and covering constraints. We extend the above main result by additionally handling a matroid independence constraints as well as finding (approximate) pareto set optimal solutions when multiple submodular objectives are present. Finally, we propose a novel and purely combinatorial dynamic programming approach that can be applied to several special cases of the problem yielding not only {\em deterministic} but also considerably faster algorithms. For example, for the well studied special case of only packing constraints (Kulik {\em et. al.} [Math. Oper. Res. `13] and Chekuri {\em et. al.} [FOCS `10]), we are able to present the first deterministic non-trivial approximation algorithm. We believe our new combinatorial approach might be of independent interest.

cs.DS↗

arXiv2018

Rational Recurrences

Despite the tremendous empirical success of neural models in natural language processing, many of them lack the strong intuitions that accompany classical machine learning approaches. Recently, connections have been shown between convolutional neural networks (CNNs) and weighted finite state automata (WFSAs), leading to new interpretations and insights. In this work, we show that some recurrent neural networks also share this connection to WFSAs. We characterize this connection formally, defining rational recurrences to be recurrent hidden state update functions that can be written as the Forward calculation of a finite set of WFSAs. We show that several recent neural models use rational recurrences. Our analysis provides a fresh view of these models and facilitates devising new neural architectures that draw inspiration from WFSAs. We present one such model, which performs better than two recent baselines on language modeling and text classification. Our results demonstrate that transferring intuitions from classical models like WFSAs can be an effective approach to designing and understanding neural models.

cs.CL↗

arXiv2018

SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference

Given a partial description like "she opened the hood of the car," humans can reason about the situation and anticipate what might come next ("then, she examined the engine"). In this paper, we introduce the task of grounded commonsense inference, unifying natural language inference and commonsense reasoning. We present SWAG, a new dataset with 113k multiple choice questions about a rich spectrum of grounded situations. To address the recurring challenges of the annotation artifacts and human biases found in many existing datasets, we propose Adversarial Filtering (AF), a novel procedure that constructs a de-biased dataset by iteratively training an ensemble of stylistic classifiers, and using them to filter the data. To account for the aggressive adversarial filtering, we use state-of-the-art language models to massively oversample a diverse set of potential counterfactuals. Empirical results demonstrate that while humans can solve the resulting inference problems with high accuracy (88%), various competitive models struggle on our task. We provide comprehensive analysis that indicates significant opportunities for future research.

cs.CL↗

arXiv2018

SoPa: Bridging CNNs, RNNs, and Weighted Finite-State Machines

Recurrent and convolutional neural networks comprise two distinct families of models that have proven to be useful for encoding natural language utterances. In this paper we present SoPa, a new model that aims to bridge these two approaches. SoPa combines neural representation learning with weighted finite-state automata (WFSAs) to learn a soft version of traditional surface patterns. We show that SoPa is an extension of a one-layer CNN, and that such CNNs are equivalent to a restricted version of SoPa, and accordingly, to a restricted form of WFSA. Empirically, on three text classification tasks, SoPa is comparable or better than both a BiLSTM (RNN) baseline and a CNN baseline, and is particularly useful in small data settings.

cs.CL↗

arXiv2018

A Dataset of Peer Reviews (PeerRead): Collection, Insights and NLP Applications

Peer reviewing is a central component in the scientific publishing process. We present the first public dataset of scientific peer reviews available for research purposes (PeerRead v1) providing an opportunity to study this important artifact. The dataset consists of 14.7K paper drafts and the corresponding accept/reject decisions in top-tier venues including ACL, NIPS and ICLR. The dataset also includes 10.7K textual peer reviews written by experts for a subset of the papers. We describe the data collection process and report interesting observed phenomena in the peer reviews. We also propose two novel NLP tasks based on this dataset and provide simple baseline models. In the first task, we show that simple models can predict whether a paper is accepted with up to 21% error reduction compared to the majority baseline. In the second task, we predict the numerical scores of review aspects and show that simple models can outperform the mean baseline for aspects with high variance such as 'originality' and 'impact'.

cs.CL↗

arXiv2018

Annotation Artifacts in Natural Language Inference Data

Large-scale datasets for natural language inference are created by presenting crowd workers with a sentence (premise), and asking them to generate three new sentences (hypotheses) that it entails, contradicts, or is logically neutral with respect to. We show that, in a significant portion of such data, this protocol leaves clues that make it possible to identify the label by looking only at the hypothesis, without observing the premise. Specifically, we show that a simple text categorization model can correctly classify the hypothesis alone in about 67% of SNLI (Bowman et. al, 2015) and 53% of MultiNLI (Williams et. al, 2017). Our analysis reveals that specific linguistic phenomena such as negation and vagueness are highly correlated with certain inference classes. Our findings suggest that the success of natural language inference models to date has been overestimated, and that the task remains a hard open problem.

cs.CL↗

arXiv2017

The Effect of Different Writing Tasks on Linguistic Style: A Case Study of the ROC Story Cloze Task

A writer's style depends not just on personal traits but also on her intent and mental state. In this paper, we show how variants of the same writing task can lead to measurable differences in writing style. We present a case study based on the story cloze task (Mostafazadeh et al., 2016a), where annotators were assigned similar writing tasks with different constraints: (1) writing an entire story, (2) adding a story ending for a given story context, and (3) adding an incoherent ending to a story. We show that a simple linear classifier informed by stylistic features is able to successfully distinguish among the three cases, without even looking at the story context. In addition, combining our stylistic features with language model predictions reaches state of the art performance on the story cloze challenge. Our results demonstrate that different task framings can dramatically affect the way people write.

cs.CL↗

arXiv2017

Automatic Selection of Context Configurations for Improved Class-Specific Word Representations

This paper is concerned with identifying contexts useful for training word representation models for different word classes such as adjectives (A), verbs (V), and nouns (N). We introduce a simple yet effective framework for an automatic selection of class-specific context configurations. We construct a context configuration space based on universal dependency relations between words, and efficiently search this space with an adapted beam search algorithm. In word similarity tasks for each word class, we show that our framework is both effective and efficient. Particularly, it improves the Spearman's rho correlation with human scores on SimLex-999 over the best previously proposed class-specific contexts by 6 (A), 6 (V) and 5 (N) rho points. With our selected context configurations, we train on only 14% (A), 26.2% (V), and 33.6% (N) of all dependency-based contexts, resulting in a reduced training time. Our results generalise: we show that the configurations our algorithm learns for one English training setup outperform previously proposed context types in another training setup for English. Moreover, basing the configuration space on universal dependencies, it is possible to transfer the learned configurations to German and Italian. We also demonstrate improved per-class results over other context types in these two languages.

cs.CL↗

arXiv2017

Local Guarantees in Graph Cuts and Clustering

Correlation Clustering is an elegant model that captures fundamental graph cut problems such as Min $s-t$ Cut, Multiway Cut, and Multicut, extensively studied in combinatorial optimization. Here, we are given a graph with edges labeled $+$ or $-$ and the goal is to produce a clustering that agrees with the labels as much as possible: $+$ edges within clusters and $-$ edges across clusters. The classical approach towards Correlation Clustering (and other graph cut problems) is to optimize a global objective. We depart from this and study local objectives: minimizing the maximum number of disagreements for edges incident on a single node, and the analogous max min agreements objective. This naturally gives rise to a family of basic min-max graph cut problems. A prototypical representative is Min Max $s-t$ Cut: find an $s-t$ cut minimizing the largest number of cut edges incident on any node. We present the following results: $(1)$ an $O(\sqrt{n})$-approximation for the problem of minimizing the maximum total weight of disagreement edges incident on any node (thus providing the first known approximation for the above family of min-max graph cut problems), $(2)$ a remarkably simple $7$-approximation for minimizing local disagreements in complete graphs (improving upon the previous best known approximation of $48$), and $(3)$ a $1/(2+\varepsilon)$-approximation for maximizing the minimum total weight of agreement edges incident on any node, hence improving upon the $1/(4+\varepsilon)$-approximation that follows from the study of approximate pure Nash equilibria in cut and party affiliation games.

cs.DS↗

arXiv2015

Online Submodular Maximization with Preemption

Submodular function maximization has been studied extensively in recent years under various constraints and models. The problem plays a major role in various disciplines. We study a natural online variant of this problem in which elements arrive one-by-one and the algorithm has to maintain a solution obeying certain constraints at all times. Upon arrival of an element, the algorithm has to decide whether to accept the element into its solution and may preempt previously chosen elements. The goal is to maximize a submodular function over the set of elements in the solution. We study two special cases of this general problem and derive upper and lower bounds on the competitive ratio. Specifically, we design a $1/e$-competitive algorithm for the unconstrained case in which the algorithm may hold any subset of the elements, and constant competitive ratio algorithms for the case where the algorithm may hold at most $k$ elements in its solution.

cs.DS↗

arXiv2014

Comparing Apples and Oranges: Query Tradeoff in Submodular Maximization

Fast algorithms for submodular maximization problems have a vast potential use in applicative settings, such as machine learning, social networks, and economics. Though fast algorithms were known for some special cases, only recently Badanidiyuru and Vondrák (2014) were the first to explicitly look for such algorithms in the general case of maximizing a monotone submodular function subject to a matroid independence constraint. The algorithm of Badanidiyuru and Vondrák matches the best possible approximation guarantee, while trying to reduce the number of value oracle queries the algorithm performs. Our main result is a new algorithm for this general case which establishes a surprising tradeoff between two seemingly unrelated quantities: the number of value oracle queries and the number of matroid independence queries performed by the algorithm. Specifically, one can decrease the former by increasing the latter and vice versa, while maintaining the best possible approximation guarantee. Such a tradeoff is very useful since various applications might incur significantly different costs in querying the value and matroid independence oracles. Furthermore, in case the rank of the matroid is $O(n^c)$, where $n$ is the size of the ground set and $c$ is an absolute constant smaller than $1$, the total number of oracle queries our algorithm uses can be made to have a smaller magnitude compared to that needed by Badanidiyuru and Vondrák. We also provide even faster algorithms for the well studied special cases of a cardinality constraint and a partition matroid independence constraint, both of which capture many real-world applications and have been widely studied both theorically and in practice.

cs.DS↗

arXiv2013

On the Approximation of Submodular Functions

Submodular functions are a fundamental object of study in combinatorial optimization, economics, machine learning, etc. and exhibit a rich combinatorial structure. Many subclasses of submodular functions have also been well studied and these subclasses widely vary in their complexity. Our motivation is to understand the relative complexity of these classes of functions. Towards this, we consider the question of how well can one class of submodular functions be approximated by another (simpler) class of submodular functions. Such approximations naturally allow algorithms designed for the simpler class to be applied to the bigger class of functions. We prove both upper and lower bounds on such approximations. Our main results are: 1. General submodular functions can be approximated by cut functions of directed graphs to a factor of $n^2/4$, which is tight. 2. General symmetric submodular functions$^{1}$ can be approximated by cut functions of undirected graphs to a factor of $n-1$, which is tight up to a constant. 3. Budgeted additive functions can be approximated by coverage functions to a factor of $e/(e-1)$, which is tight. Here $n$ is the size of the ground set on which the submodular function is defined. We also observe that prior works imply that monotone submodular functions can be approximated by coverage functions with a factor between $O(\sqrt{n} \log n)$ and $Ω(n^{1/3} /\log^2 n) $.

cs.DS↗

arXiv2012

On the Furthest Hyperplane Problem and Maximal Margin Clustering

This paper introduces the Furthest Hyperplane Problem (FHP), which is an unsupervised counterpart of Support Vector Machines. Given a set of n points in Rd, the objective is to produce the hyperplane (passing through the origin) which maximizes the separation margin, that is, the minimal distance between the hyperplane and any input point. To the best of our knowledge, this is the first paper achieving provable results regarding FHP. We provide both lower and upper bounds to this NP-hard problem. First, we give a simple randomized algorithm whose running time is n^O(1/θ^2) where θ is the optimal separation margin. We show that its exponential dependency on 1/θ^2 is tight, up to sub-polynomial factors, assuming SAT cannot be solved in sub-exponential time. Next, we give an efficient approxima- tion algorithm. For any α \in [0, 1], the algorithm produces a hyperplane whose distance from at least 1 - 5α fraction of the points is at least α times the optimal separation margin. Finally, we show that FHP does not admit a PTAS by presenting a gap preserving reduction from a particular version of the PCP theorem.

cs.CC↗

arXiv2011

Min-Max Graph Partitioning and Small Set Expansion

We study graph partitioning problems from a min-max perspective, in which an input graph on n vertices should be partitioned into k parts, and the objective is to minimize the maximum number of edges leaving a single part. The two main versions we consider are where the k parts need to be of equal-size, and where they must separate a set of k given terminals. We consider a common generalization of these two problems, and design for it an $O(\sqrt{\log n\log k})$-approximation algorithm. This improves over an $O(\log^2 n)$ approximation for the second version, and roughly $O(k\log n)$ approximation for the first version that follows from other previous work. We also give an improved O(1)-approximation algorithm for graphs that exclude any fixed minor. Our algorithm uses a new procedure for solving the Small-Set Expansion problem. In this problem, we are given a graph G and the goal is to find a non-empty set $S\subseteq V$ of size $|S| \leq ρn$ with minimum edge-expansion. We give an $O(\sqrt{\log{n}\log{(1/ρ)}})$ bicriteria approximation algorithm for the general case of Small-Set Expansion, and O(1) approximation algorithm for graphs that exclude any fixed minor.

cs.DS↗