Literature DB >> 33400900

Whence the Expected Free Energy?

Beren Millidge1, Alexander Tschantz2, Christopher L Buckley3.   

Abstract

The expected free energy (EFE) is a central quantity in the theory of active inference. It is the quantity that all active inference agents are mandated to minimize through action, and its decomposition into extrinsic and intrinsic value terms is key to the balance of exploration and exploitation that active inference agents evince. Despite its importance, the mathematical origins of this quantity and its relation to the variational free energy (VFE) remain unclear. In this letter, we investigate the origins of the EFE in detail and show that it is not simply "the free energy in the future." We present a functional that we argue is the natural extension of the VFE but actively discourages exploratory behavior, thus demonstrating that exploration does not directly follow from free energy minimization into the future. We then develop a novel objective, the free energy of the expected future (FEEF), which possesses both the epistemic component of the EFE and an intuitive mathematical grounding as the divergence between predicted and desired futures.

Year:  2021        PMID: 33400900     DOI: 10.1162/neco_a_01354

Source DB:  PubMed          Journal:  Neural Comput        ISSN: 0899-7667            Impact factor:   2.026


  7 in total

1.  Epistemic Communities under Active Inference.

Authors:  Mahault Albarracin; Daphne Demekas; Maxwell J D Ramstead; Conor Heins
Journal:  Entropy (Basel)       Date:  2022-03-29       Impact factor: 2.738

2.  The Problem of Meaning: The Free Energy Principle and Artificial Agency.

Authors:  Julian Kiverstein; Michael D Kirchhoff; Tom Froese
Journal:  Front Neurorobot       Date:  2022-06-23       Impact factor: 3.493

3.  A step-by-step tutorial on active inference and its application to empirical data.

Authors:  Ryan Smith; Karl J Friston; Christopher J Whyte
Journal:  J Math Psychol       Date:  2022-02-04       Impact factor: 1.387

Review 4.  The Free Energy Principle for Perception and Action: A Deep Learning Perspective.

Authors:  Pietro Mazzaglia; Tim Verbelen; Ozan Çatal; Bart Dhoedt
Journal:  Entropy (Basel)       Date:  2022-02-21       Impact factor: 2.524

5.  Interoception as modeling, allostasis as control.

Authors:  Eli Sennesh; Jordan Theriault; Dana Brooks; Jan-Willem van de Meent; Lisa Feldman Barrett; Karen S Quigley
Journal:  Biol Psychol       Date:  2021-12-20       Impact factor: 3.111

6.  Active Inference and Epistemic Value in Graphical Models.

Authors:  Thijs van de Laar; Magnus Koudahl; Bart van Erp; Bert de Vries
Journal:  Front Robot AI       Date:  2022-04-06

7.  On Epistemics in Expected Free Energy for Linear Gaussian State Space Models.

Authors:  Magnus T Koudahl; Wouter M Kouw; Bert de Vries
Journal:  Entropy (Basel)       Date:  2021-11-24       Impact factor: 2.524

  7 in total

北京卡尤迪生物科技股份有限公司 © 2022-2023.