Implicit Training of Energy Model for Structure Prediction

TL;DR

This paper introduces an implicit training approach for energy-based models in structure prediction, optimizing dynamic loss objectives via implicit gradients to enhance performance on complex structured outputs.

Contribution

It proposes using implicit-gradient techniques to learn dynamic loss landscapes, improving structure prediction beyond standard loss functions.

Findings

Implicitly learned dynamic loss improves accuracy.

Energy model training benefits from implicit gradient optimization.

Method outperforms traditional static loss approaches.

Abstract

Most deep learning research has focused on developing new model and training procedures. On the other hand the training objective has usually been restricted to combinations of standard losses. When the objective aligns well with the evaluation metric, this is not a major issue. However when dealing with complex structured outputs, the ideal objective can be hard to optimize and the efficacy of usual objectives as a proxy for the true objective can be questionable. In this work, we argue that the existing inference network based structure prediction methods ( Tu and Gimpel 2018; Tu, Pang, and Gimpel 2020) are indirectly learning to optimize a dynamic loss objective parameterized by the energy model. We then explore using implicit-gradient based technique to learn the corresponding dynamic objectives. Our experiments show that implicitly learning a dynamic loss landscape is an effective method for improving model performance in structure prediction.