Active Inference In Racket

Overview

This skill provides a complete Active Inference implementation in Racket, demonstrating Bayesian belief updating, variational free energy calculation, and expected free energy-based policy selection.

Core Algorithms

1. Belief Updating: Bayesian inference using observation likelihoods to update posterior beliefs
2. Free Energy Calculation: KL divergence between posterior beliefs and prior distribution
3. Policy Selection: Softmax action selection over expected free energy per action
4. Perception-Action Loop: Iterative sense → infer → act cycle with generative model

Key Files

• active_inference.rkt — Source implementation
• run.sh — Execution script (handles compilation if needed)
• README.md — Usage documentation and requirements

Usage