Introducing ActiveInference.jl: A Julia Library for Simulation and Parameter Estimation with Active Inference Models.

Area of Science:

• Computational neuroscience
• Cognitive science
• Computational psychiatry

Background:

• Active inference is a framework for understanding how agents interact with their environment.
• Partially Observable Markov Decision Processes (POMDPs) are used to model decision-making under uncertainty.
• Existing tools for active inference, like pymdp, are primarily available in Python.

Purpose of the Study:

• To introduce ActiveInference.jl, a new software package for the Julia programming language.
• To make active inference agents with POMDP generative models accessible to the Julia research community.
• To facilitate the use of active inference models for simulation, data fitting, and model comparison.

Main Methods:

• Re-implementation of the pymdp library for Python into ActiveInference.jl for Julia.
• Ensuring compatibility with existing Julia libraries for cognitive and behavioral modeling.
• Utilizing sampling and variational methods for fitting POMDP active inference models to empirical data.

Main Results:

• ActiveInference.jl provides a straightforward way to build POMDP active inference models.
• The package enables researchers to easily fit models to observed behavior.
• Researchers can use ActiveInference.jl for simulation and model comparison.

Conclusions:

• ActiveInference.jl lowers the barrier to entry for using active inference with POMDPs in Julia.
• The package supports advanced computational modeling in cognitive science, neuroscience, and psychiatry.
• This facilitates the integration of theoretical active inference frameworks with empirical research.