Emergence of Compositional Representations in Restricted Boltzmann Machines

TL;DR

This paper investigates the conditions under which Restricted Boltzmann Machines develop compositional representations, combining theoretical analysis and experiments on MNIST to understand their feature extraction capabilities.

Contribution

It identifies key structural conditions enabling RBMs to learn compositional features, supported by replica analysis and empirical training results.

Findings

RBMs can operate in a compositional phase under specific conditions.

Replica analysis confirms the theoretical conditions for compositionality.

Trained RBMs on MNIST exhibit compositional feature representations.

Abstract

Extracting automatically the complex set of features composing real high-dimensional data is crucial for achieving high performance in machine--learning tasks. Restricted Boltzmann Machines (RBM) are empirically known to be efficient for this purpose, and to be able to generate distributed and graded representations of the data. We characterize the structural conditions (sparsity of the weights, low effective temperature, nonlinearities in the activation functions of hidden units, and adaptation of fields maintaining the activity in the visible layer) allowing RBM to operate in such a compositional phase. Evidence is provided by the replica analysis of an adequate statistical ensemble of random RBMs and by RBM trained on the handwritten digits dataset MNIST.