Structured Prediction via the Extragradient Method

By B. Taskar, Simon Lacoste-Julien, Michael I. Jordan. 2005

63 citations recorded by the source corpus.

Subfields: Information Extraction, Mathematical Optimization

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Information Extraction

In Mathematical Optimization

Citations in the corpus

Maximum margin planning (2006, background): “(Taskar et al., 2006) The latter is applicable where inference may be written as a linear program and is also able to achieve linear convergence rates.”
A Discourse Commitment-Based Framework for Recognizing Textual Entailment (2007, methodology): “Following Commitment Extraction, we used an word alignment technique ﬁrst introduced in (Taskar et al., 2005b) in order to select the commitment extracted from t (henceforth, c t ) which represents the best alignment for each of the commitments extracted from h (henceforth, c h ).”
Learning Gaussian Conditional Random Fields for Low-Level Vision (2007, methodology): “This type of training has also been used for discrete-valued graphical models by Collins [4], Taskar et al. [20, 19], and Altun et al. [1].”
Cutting-plane training of structural SVMs (2009, methodology): “Similar restrictions also apply to the extragradient method (Taskar et al. 2005), which applies only to problems where subgradients of the QP can be computed via a convex real relaxation, as well as exponentiated gradient methods (Bartlett et al. 2004; Globerson et al. 2007), which require the…”
Subgradient Methods for Structured Prediction (2007, background): “Recently, (Taskar et al., 2006) have investigated saddle-point methods for optimization and have succeeded in efficiently solving several problems that would have otherwise had intractable memory requirements.”
Learning to search: Functional gradient techniques for imitation learning (2009, methodology): “If we restrict Gi (which is formally interpreted as the (exponentially large) set of classes in a structured prediction problem ( Taskar et al. 2006 )), to contain only two elements and choose the loss function to be the Hamming loss, then the convex program reduces to the one typically seen in the ”
Structured Prediction, Dual Extragradient and Bregman Projections (2006, methodology): “In our previous work (Taskar et al., 2006), we used a related method, the extragradient method due to Korpelevich (1976).”
Learning optimized MAP estimates in continuously-valued MRF models (2009, background): “have shown that for certain types of energy and loss functions, the learning task in Equation 1 can be accomplished with convex optimization algorithms [16, 17].”