Supervised structure learning

TL;DR

This paper introduces a Bayesian approach to structure learning that uses expected free energy to guide model discovery and data assimilation, demonstrated through image classification and dynamic model discovery tasks.

Contribution

It presents a novel scheme that employs priors on model selection based on expected free energy, enabling autodidactic construction of generative models for disentangling latent structures.

Findings

Successful image classification on MNIST

Disentangled latent structures in dynamic models

Effective discovery of factorial state representations

Abstract

This paper concerns structure learning or discovery of discrete generative models. It focuses on Bayesian model selection and the assimilation of training data or content, with a special emphasis on the order in which data are ingested. A key move - in the ensuing schemes - is to place priors on the selection of models, based upon expected free energy. In this setting, expected free energy reduces to a constrained mutual information, where the constraints inherit from priors over outcomes (i.e., preferred outcomes). The resulting scheme is first used to perform image classification on the MNIST dataset to illustrate the basic idea, and then tested on a more challenging problem of discovering models with dynamics, using a simple sprite-based visual disentanglement paradigm and the Tower of Hanoi (cf., blocks world) problem. In these examples, generative models are constructed autodidactically to recover (i.e., disentangle) the factorial structure of latent states - and their characteristic paths or dynamics.