Perception-Distortion Trade-off with Restricted Boltzmann Machines

TL;DR

This paper proposes a novel method for using Restricted Boltzmann Machines in missing data inference by linearizing the energy function, and explores the perception-distortion trade-off in this context.

Contribution

It introduces a new RBM-based inference procedure for incomplete data and analyzes the perception-distortion trade-off in this setting.

Findings

The proposed RBM procedure outperforms existing methods on missing data tasks.

The perception-distortion trade-off is significant in incomplete data reconstruction.

Linearization of the energy function improves inference accuracy.

Abstract

In this work, we introduce a new procedure for applying Restricted Boltzmann Machines (RBMs) to missing data inference tasks, based on linearization of the effective energy function governing the distribution of observations. We compare the performance of our proposed procedure with those obtained using existing reconstruction procedures trained on incomplete data. We place these performance comparisons within the context of the perception-distortion trade-off observed in other data reconstruction tasks, which has, until now, remained unexplored in tasks relying on incomplete training data.