Linear Genetic Programming
Rust LGP framework with Q-Learning integration
Overview
Cargo workspace implementing linear genetic programming as a trait-based Rust library with a CLI for experiment automation. Supports RL environments (CartPole, MountainCar) and classification tasks (Iris) with Q-Learning integration, Rayon-powered parallel evaluation, and built-in hyperparameter search.
Demo
Baseline
Crossover
Mutation
Full Pipeline
Features
Trait-Based Architecture
Swap genetic operators, fitness functions, and selection via Rust traits.
Parallel Evaluation
Rayon-powered parallel fitness evaluation distributing work across all CPU cores.
Q-Learning Hybrid
RL layer learns register-action mappings on top of evolved programs.
Experiment CLI
Full pipeline: hyperparameter search, batch runs, and analysis in one tool.
Structured Tracing
Multi-level structured tracing with JSON output for log aggregation systems.
Hyperparameter Search
Built-in random search with parallel Rayon evaluation and optimal config export.