Tomographic Auto-Encoder: Unsupervised Bayesian Recovery of Corrupted Data

TL;DR

The paper introduces a probabilistic auto-encoder that effectively recovers corrupted data by modeling uncertainty, outperforming existing methods in imputation, denoising, and improving downstream classification accuracy.

Contribution

It presents a novel reduced entropy approximate inference method for rich posterior recovery in unsupervised data restoration tasks.

Findings

Superior imputation and denoising performance over existing variational methods

Higher classification accuracy after data imputation

Effective exploration of data manifold and uncertainty characterization

Abstract

We propose a new probabilistic method for unsupervised recovery of corrupted data. Given a large ensemble of degraded samples, our method recovers accurate posteriors of clean values, allowing the exploration of the manifold of possible reconstructed data and hence characterising the underlying uncertainty. In this setting, direct application of classical variational methods often gives rise to collapsed densities that do not adequately explore the solution space. Instead, we derive our novel reduced entropy condition approximate inference method that results in rich posteriors. We test our model in a data recovery task under the common setting of missing values and noise, demonstrating superior performance to existing variational methods for imputation and de-noising with different real data sets. We further show higher classification accuracy after imputation, proving the advantage of propagating uncertainty to downstream tasks with our model.