Now showing items 21-39 of 39

    • Low-Rank Tensors for Scoring Dependency Structures 

      Lei, Tao; Zhang, Yuan; Barzilay, Regina; Jaakkola, Tommi S. (Association for Computational Linguistics, 2014-06)
      Accurate scoring of syntactic structures such as head-modifier arcs in dependency parsing typically requires rich, high-dimensional feature representations. A small subset of such features is often selected manually. This ...
    • Metric recovery from directed unweighted graphs 

      Hashimoto, Tatsunori Benjamin; Sun, Yi; Jaakkola, Tommi S (PMLR, 2015-05)
      We analyze directed, unweighted graphs obtained from x[subscript i] ∈ R[superscript d] by connecting vertex i to j iff |x[subscript i] − x[subscript j]| < ε(x[subscript i]). Examples of such graphs include k-nearest neighbor ...
    • More data means less inference: A pseudo-max approach to structured learning 

      Sontag, David; Meshi, Ofer; Jaakkola, Tommi S.; Globerson, Amir (Neural Information Processing Systems Foundation, 2010-12)
      The problem of learning to predict structured labels is of key importance in many applications. However, for general graph structure both learning and inference in this setting are intractable. Here we show that it is ...
    • On dual decomposition and linear programming relaxations for natural language processing 

      Rush, Alexander Matthew; Sontag, David Alexander; Collins, Michael; Jaakkola, Tommi S. (Association for Computational Linguistics, 2010-10)
      This paper introduces dual decomposition as a framework for deriving inference algorithms for NLP problems. The approach relies on standard dynamic-programming algorithms as oracle solvers for sub-problems, together ...
    • On measure concentration of random maximum a-posteriori perturbations 

      Orabona, Francesco; Hazan, Tamir; Sarwate, Anand D.; Jaakkola, Tommi S. (Association for Computing Machinery (ACM), 2014)
      The maximum a-posteriori (MAP) perturbation framework has emerged as a useful approach for inference and learning in high dimensional complex models. By maximizing a randomly perturbed potential function, MAP perturbations ...
    • On sampling from the Gibbs distribution with random maximum a-posteriori perturbations 

      Hazan, Tamir; Maji, Subhransu; Jaakkola, Tommi S. (Neural Information Processing Systems, 2013)
      In this paper we describe how MAP inference can be used to sample efficiently from Gibbs distributions. Specifically, we provide means for drawing either approximate or unbiased samples from Gibbs' distributions by introducing ...
    • Predicting organic reaction outcomes with weisfeiler-lehman network 

      Jin, Wengong; Coley, Connor Wilson; Barzilay, Regina; Jaakkola, Tommi S (Neural Information Processing Systems Foundation, Inc., 2017-12)
      The prediction of organic reaction outcomes is a fundamental problem in computational chemistry. Since a reaction may involve hundreds of atoms, fully exploring the space of possible transformations is intractable. The ...
    • Principal differences analysis: Interpretable characterization of differences between distributions 

      Mueller, Jonas Weylin; Jaakkola, Tommi S (Neural Information Processing Systems Foundation, Inc., 2015-12)
      We introduce principal differences analysis (PDA) for analyzing differences between high-dimensional distributions. The method operates by finding the projection that maximizes the Wasserstein divergence between the resulting ...
    • Rationalizing Neural Predictions 

      Lei, Tao; Barzilay, Regina; Jaakkola, Tommi S (Association for Computational Linguistics (ACL), 2016-11)
      Prediction without justification has limited applicability. As a remedy, we learn to extract pieces of input text as justifications – rationales – that are tailored to be short and coherent, yet sufficient for making the ...
    • Rethinking Cooperative Rationalization: Introspective Extraction and Complement Control 

      Yu, Mo; Chang, Shiyu; Zhang, Yang; Jaakkola, Tommi S (Association for Computational Linguistics, 2019-11)
      Selective rationalization has become a common mechanism to ensure that predictive models reveal how they use any available features. The selection may be soft or hard, and identifies a subset of input features relevant for ...
    • Solving graph compression via optimal transport 

      Garg, Vikas; Jaakkola, Tommi S (Morgan Kaufmann Publishers, 2019-12)
      We propose a new approach to graph compression by appeal to optimal transport. The transport problem is seeded with prior information about node importance, attributes, and edges in the graph. The transport formulation can ...
    • Steps to Excellence: Simple Inference with Refined Scoring of Dependency Trees 

      Zhang, Yuan; Lei, Tao; Barzilay, Regina; Jaakkola, Tommi S.; Globerson, Amir (Association for Computational Linguistics, 2014-06)
      Much of the recent work on dependency parsing has been focused on solving inherent combinatorial problems associated with rich scoring functions. In contrast, we demonstrate that highly expressive scoring functions can be ...
    • Tight certificates of adversarial robustness for randomly smoothed classifiers 

      Lee, Guang-He; Yuan, Yang; Jaakkola, Tommi S (2020-02)
      Strong theoretical guarantees of robustness can be given for ensembles of classifiers generated by input randomization. Specifically, an `2 bounded adversary cannot alter the ensemble prediction generated by an additive ...
    • Towards optimal transport with global invariances 

      Alvarez Melis, David; Jegelka, Stefanie Sabrina; Jaakkola, Tommi S (JMLR, 2019-02)
      Many problems in machine learning involve calculating correspondences between sets of objects, such as point clouds or images. Discrete optimal transport provides a natural and successful approach to such tasks whenever ...
    • Tree block coordinate descent for map in graphical models 

      Sontag, David Alexander; Jaakkola, Tommi S. (Journal of Machine Learning Research, 2009-04)
      A number of linear programming relaxations have been proposed for finding most likely settings of the variables (MAP) in large probabilistic models. The relaxations are often succinctly expressed in the dual and reduce to ...
    • Two-sided exponential concentration bounds for Bayes error rate and Shannon entropy 

      Honorio, Jean; Jaakkola, Tommi S. (Association for Computing Machinery (ACM), 2013)
      We provide a method that approximates the Bayes error rate and the Shannon entropy with high probability. The Bayes error rate approximation makes possible to build a classifier that polynomially approaches Bayes error ...
    • A unified framework for consistency of regularized loss minimizers 

      Honorio, Jean; Jaakkola, Tommi S. (Association for Computing Machinery (ACM), 2014)
      We characterize a family of regularized loss minimization problems that satisfy three properties: scaled uniform convergence, super-norm regularization, and norm-loss monotonicity. We show several theoretical guarantees ...
    • An unsupervised method for uncovering morphological chains 

      Narasimhan, Karthik Rajagopal; Barzilay, Regina; Jaakkola, Tommi S. (Association for Computational Linguistics, 2015-03)
      Most state-of-the-art systems today produce morphological analysis based only on orthographic patterns. In contrast, we propose a model for unsupervised morphological analysis that integrates orthographic and semantic views ...
    • Validation and refinement of gene-regulatory pathways on a network of physical interactions 

      Yeang, Chen-Hsiang, 1969-; Mak, Craig; McCuine, Scott; Workman, Christopher; Ideker, Trey; e.a. (BioMed Central Ltd, 2005-07)
      As genome-scale measurements lead to increasingly complex models of gene regulation, systematic approaches are needed to validate and refine these models. Towards this goal, we describe an automated procedure for prioritizing ...