Supervised structure learning

Karl J Friston; Lancelot Da Costa; Alexander Tschantz; Alex Kiefer; Tommaso Salvatori; Victorita Neacsu; Magnus Koudahl; Conor Heins; Noor Sajid; Dimitrije Markovic; Thomas Parr; Tim Verbelen; Christopher L Buckley

doi:10.1016/j.biopsycho.2024.108891

Supervised structure learning

Biol Psychol. 2024 Oct 19:193:108891. doi: 10.1016/j.biopsycho.2024.108891. Online ahead of print.

Authors

Affiliations

¹ Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, UK; VERSES AI Research Lab, Los Angeles, CA, 90016, USA.
² Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, UK; VERSES AI Research Lab, Los Angeles, CA, 90016, USA; Department of Mathematics, Imperial College London, UK.
³ VERSES AI Research Lab, Los Angeles, CA, 90016, USA; School of Engineering and Informatics, University of Sussex, Brighton, UK. Electronic address: tschantz.alec@gmail.com.
⁴ VERSES AI Research Lab, Los Angeles, CA, 90016, USA.
⁵ Wellcome Trust Centre for Neuroimaging, Institute of Neurology, University College London, UK.
⁶ Chair of Cognitive Computational Neuroscience, Technische Universität Dresden, Dresden, Germany.
⁷ Nuffield Department of Clinical Neurosciences, University of Oxford, UK.
⁸ VERSES AI Research Lab, Los Angeles, CA, 90016, USA; School of Engineering and Informatics, University of Sussex, Brighton, UK.

PMID: 39433209
DOI: 10.1016/j.biopsycho.2024.108891

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.

Keywords: Active inference; Active learning; Bayesian model selection; Disentanglement; Expected free energy; Planning as inference; Structure learning.