EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

From pixels to planning: scale-free active inference

Karl Friston, Conor Heins, Tim Verbelen, Lancelot Da Costa, Tommaso Salvatori, Dimitrije Markovic, Alexander Tschantz, Magnus Koudahl, Christopher Buckley, Thomas Parr

Frontiers in Network Physiology · 18.06.2025 · 5

DOI: 10.3389/fnetp.2025.1521963 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Network Physiology, online erschienen am 18.06.2025. Verfasst von Karl Friston und 9 weiteren Autor:innen. Beteiligte Einrichtung: Queen Square Institute of Neurology, University College London, London, United Kingdom. Die Arbeit wurde bisher 16-mal zitiert (Crossref) und nennt 129 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Wellcome Trust.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper describes a discrete state-space model and accompanying methods for generative modeling. This model generalizes partially observed Markov decision processes to include paths as latent variables, rendering it suitable for active inference and learning in a dynamic setting. Specifically, we consider deep or hierarchical forms using the renormalization group. The ensuing renormalizing generative models (RGM) can be regarded as discrete homologs of deep convolutional neural networks or continuous state-space models in generalized coordinates of motion. By construction, these scale-invariant models can be used to learn compositionality over space and time, furnishing models of paths or orbits: that is, events of increasing temporal depth and itinerancy. This technical note illustrates the automatic discovery, learning, and deployment of RGMs using a series of applications. We start with image classification and then consider the compression and generation of movies and music. Finally, we apply the same variational principles to the learning of Atari-like games.

Schlagwörter

Stichwörter der Autor:innen

Bayesian model selectionactive inferenceactive learningcompressionnetwork-physiologyrenormalization groupstructure learning

Autor:innen und Institutionen

  • Queen Square Institute of Neurology, University College London, London, United Kingdom.
  • VERSES Research Lab, Los Angeles, CA, United States.
  • Cognitive Computational Neuroscience, Technische Universität, Dresden, Germany.
  • Department of Informatics, University of Sussex, Brighton, United Kingdom.
  • Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom.

Förderung

  • Wellcome Trust

Zitieren

APA

Friston, K., Heins, C., Verbelen, T., Da Costa, L., Salvatori, T., Markovic, D., Tschantz, A., Koudahl, M., Buckley, C., & Parr, T. (2025). From pixels to planning: scale-free active inference. Frontiers in Network Physiology, 5. https://doi.org/10.3389/fnetp.2025.1521963

Vancouver

Friston K, Heins C, Verbelen T, Da Costa L, Salvatori T, Markovic D, et al. From pixels to planning: scale-free active inference. Frontiers in Network Physiology. 2025;5. doi:10.3389/fnetp.2025.1521963

BibTeX
@article{Friston2025,
  author = {Friston, Karl and Heins, Conor and Verbelen, Tim and Da Costa, Lancelot and Salvatori, Tommaso and Markovic, Dimitrije and Tschantz, Alexander and Koudahl, Magnus and Buckley, Christopher and Parr, Thomas},
  title = {From pixels to planning: scale-free active inference},
  journal = {Frontiers in Network Physiology},
  year = {2025},
  volume = {5},
  doi = {10.3389/fnetp.2025.1521963}
}
Kostenlos starten →