A Truncated EM Approach for Spike-and-Slab Sparse Coding

TL;DR

This paper introduces a truncated EM method for spike-and-slab sparse coding, demonstrating improved inference and learning over traditional approaches, especially in source separation and image denoising tasks with high noise levels.

Contribution

The paper proposes a novel truncated EM algorithm for spike-and-slab sparse coding, outperforming standard variational methods in source separation and denoising benchmarks.

Findings

Both approaches improve state-of-the-art in benchmarks.

Truncated EM outperforms factored variational in source separation.

Truncated EM improves performance at higher noise levels.

Abstract

We study inference and learning based on a sparse coding model with `spike-and-slab' prior. As in standard sparse coding, the model used assumes independent latent sources that linearly combine to generate data points. However, instead of using a standard sparse prior such as a Laplace distribution, we study the application of a more flexible `spike-and-slab' distribution which models the absence or presence of a source's contribution independently of its strength if it contributes. We investigate two approaches to optimize the parameters of spike-and-slab sparse coding: a novel truncated EM approach and, for comparison, an approach based on standard factored variational distributions. The truncated approach can be regarded as a variational approach with truncated posteriors as variational distributions. In applications to source separation we find that both approaches improve the state-of-the-art in a number of standard benchmarks, which argues for the use of `spike-and-slab' priors for the corresponding data domains. Furthermore, we find that the truncated EM approach improves on the standard factored approach in source separation tasks$-$which hints to biases introduced by assuming posterior independence in the factored variational approach. Likewise, on a standard benchmark for image denoising, we find that the truncated EM approach improves on the factored variational approach. While the performance of the factored approach saturates with increasing numbers of hidden dimensions, the performance of the truncated approach improves the state-of-the-art for higher noise levels.