Statistical-mechanical study of deep Boltzmann machine given weight parameters after training by singular value decomposition

TL;DR

This paper uses statistical mechanics to analyze deep Boltzmann machines, revealing how weight correlations and layer numbers influence their generative performance.

Contribution

It provides a theoretical phase diagram for DBMs based on weight properties and layer configurations using the replica method.

Findings

Correlation between hidden layer weights is crucial for performance.

Increasing layers can negatively impact generative ability if weight correlation is low.

Derived phase diagram depends on layer and unit numbers.

Abstract

Deep learning methods relying on multi-layered networks have been actively studied in a wide range of fields in recent years, and deep Boltzmann machines(DBMs) is one of them. In this study, a model of DBMs with some properites of weight parameters obtained by learning is studied theoretically by a statistical-mechanical approach based on the replica method. The phases characterizing the role of DBMs as a generator and their phase diagram are derived, depending meaningfully on the numbers of layers and units in each layer. It is found, in particular, that the correlation between the weight parameters in the hidden layers plays an essential role and that an increase in the layer number has a negative effect on DBM's performance as a generator when the correlation is smaller than a certain threshold value.