On Optimality Conditions for Auto-Encoder Signal Recovery

TL;DR

This paper explores the conditions under which auto-encoders can accurately recover true hidden signals, linking them to concepts in sparse coding and compressed sensing, and demonstrates their effectiveness in recovering data dictionaries.

Contribution

It establishes theoretical conditions for auto-encoder signal recovery and empirically shows their ability to recover data generating dictionaries from samples.

Findings

True hidden representations can be approximately recovered under certain incoherence and bias conditions.

Recovery accuracy improves with increased sparsity in signals.

Auto-encoders can recover data dictionaries from samples.

Abstract

Auto-Encoders are unsupervised models that aim to learn patterns from observed data by minimizing a reconstruction cost. The useful representations learned are often found to be sparse and distributed. On the other hand, compressed sensing and sparse coding assume a data generating process, where the observed data is generated from some true latent signal source, and try to recover the corresponding signal from measurements. Looking at auto-encoders from this \textit{signal recovery perspective} enables us to have a more coherent view of these techniques. In this paper, in particular, we show that the \textit{true} hidden representation can be approximately recovered if the weight matrices are highly incoherent with unit $ \ell^{2} $ row length and the bias vectors takes the value (approximately) equal to the negative of the data mean. The recovery also becomes more and more accurate as the sparsity in hidden signals increases. Additionally, we empirically demonstrate that auto-encoders are capable of recovering the data generating dictionary when only data samples are given.