Learning Approximate Inference Networks for Structured Prediction

TL;DR

This paper introduces a neural network-based inference method for structured prediction that replaces gradient descent, achieving significant speed-ups and improved accuracy in various structured prediction tasks.

Contribution

It proposes a neural inference network trained to approximate structured argmax inference, replacing gradient-based methods for faster and more accurate structured prediction.

Findings

10-60x speed-up in multi-label classification

Comparable speed and accuracy to exact inference in sequence labeling

Improved accuracy with label language model for long-distance dependencies

Abstract

Structured prediction energy networks (SPENs; Belanger & McCallum 2016) use neural network architectures to define energy functions that can capture arbitrary dependencies among parts of structured outputs. Prior work used gradient descent for inference, relaxing the structured output to a set of continuous variables and then optimizing the energy with respect to them. We replace this use of gradient descent with a neural network trained to approximate structured argmax inference. This "inference network" outputs continuous values that we treat as the output structure. We develop large-margin training criteria for joint training of the structured energy function and inference network. On multi-label classification we report speed-ups of 10-60x compared to (Belanger et al, 2017) while also improving accuracy. For sequence labeling with simple structured energies, our approach performs comparably to exact inference while being much faster at test time. We then demonstrate improved accuracy by augmenting the energy with a "label language model" that scores entire output label sequences, showing it can improve handling of long-distance dependencies in part-of-speech tagging. Finally, we show how inference networks can replace dynamic programming for test-time inference in conditional random fields, suggestive for their general use for fast inference in structured settings.