ForneyLab: A Toolbox for Biologically Plausible Free Energy Minimization in Dynamic Neural Models


The free energy principle (FEP) claims that self-organization in biological agents is driven by variational free energy (FE) minimization in a generative probabilistic model of the agent’s environment. We have developed ForneyLab as a freely available toolbox for FE minimization by variational message passing in probabilistic dynamic models. We introduce the FFG formalism and ForneyLab by presenting two example applications that are well-known to FEP researchers. With this toolbox we aim to lubricate the execution of research projects on biologically plausible probabilistic modeling.

2018 Conference on Complex Systems