Open AccessProceedings Article
Implicit Differentiation by Perturbation
Justin Domke
- 06 Dec 2010
- Vol. 23, pp 523-531
TL;DR: A simple and efficient finite difference method for implicit differentiation of marginal inference results in discrete graphical models by showing that the derivatives of this loss with respect to model parameters can be obtained by running the inference procedure twice, on slightly perturbed model parameters.
read more
Abstract: This paper proposes a simple and efficient finite difference method for implicit differentiation of marginal inference results in discrete graphical models. Given an arbitrary loss function, defined on marginals, we show that the derivatives of this loss with respect to model parameters can be obtained by running the inference procedure twice, on slightly perturbed model parameters. This method can be used with approximate inference, with a loss function over approximate marginals. Convenient choices of loss functions make it practical to fit graphical models with hidden variables, high treewidth and/or model misspecification.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Citations
•Proceedings Article
Cardinality Restricted Boltzmann Machines
Kevin Swersky,Ilya Sutskever,Daniel Tarlow,Richard S. Zemel,Ruslan Salakhutdinov,Ryan P. Adams +5 more
- 03 Dec 2012
TL;DR: It is shown that a dynamic programming algorithm can be used to implement exact sparsity in the RBM's hidden units and how to pass derivatives through the resulting posterior marginals, which makes it possible to fine-tune a pre-trained neural network with sparse hidden layers.
Approximation-Aware Dependency Parsing by Belief Propagation
TL;DR: This work trains the fast dependency parser of Smith and Eisner (2008) for improved accuracy by treating the entire parser (approximations and all) as a differentiable circuit, as others have done for loopy CRFs.
•Proceedings Article
Minimum-Risk Training of Approximate CRF-Based NLP Systems
Veselin Stoyanov,Jason Eisner +1 more
- 03 Jun 2012
TL;DR: This work applies the Stoyanov et al. method to three NLP problems, showing that using more complex CRFs leads to improved performance, and that minimum-risk training learns more accurate models.
•Posted Content
Gradient Estimation with Stochastic Softmax Tricks.
TL;DR: The authors introduce stochastic softmax tricks, which generalize the Gumbel-Softmax trick to combinatorial spaces, and design structured relaxations for subset selection, spanning trees, arborescences, and others.
28
References
•Book
Graphical Models, Exponential Families, and Variational Inference
Martin J. Wainwright,Michael I. Jordan +1 more
- 16 Dec 2008
TL;DR: The variational approach provides a complementary alternative to Markov chain Monte Carlo as a general source of approximation methods for inference in large-scale statistical models.
Constructing free-energy approximations and generalized belief propagation algorithms
TL;DR: This work explains how to obtain region-based free energy approximations that improve the Bethe approximation, and corresponding generalized belief propagation (GBP) algorithms, and describes empirical results showing that GBP can significantly outperform BP.
Constructing free-energy approximations and generalized belief propagation algorithms
TL;DR: This work explains how to obtain region-based free energy approximations that improve the Bethe approximation, and corresponding generalized belief propagation (GBP) algorithms, and describes empirical results showing that GBP can significantly outperform BP.
1K
libDAI: A Free and Open Source C++ Library for Discrete Approximate Inference in Graphical Models
TL;DR: The software package libDAI, a free & open source C++ library that provides implementations of various exact and approximate inference methods for graphical models with discrete-valued variables, is described.
CCCP algorithms to minimize the Bethe and Kikuchi free energies: convergent alternatives to belief propagation
TL;DR: A class of discrete iterative algorithms that are provably convergent alternatives to believe propagation (BP) and generalized belief propagation (GBP) and are pointed out that have a large range of inference and learning applications.
318