RL Experiments

This repository is a workspace for running and tracking reinforcement learning (RL) experiments. It contains small prototypes, training scripts, and plots used to explore algorithms and environments. Everything is implemented from scratch.

Experiments

• CartPole: REINFORCE (Monte-Carlo policy gradient) with per-episode updates and training curves in cartpole/.
• LunarLanderContinuous: PPO (actor-critic) with GAE and clipped objective, with metrics plots in lunar_lander/.

Structure

All experiments share a common structure:

• README.md: description of the experiment
• train.py: training code and experiment-specific assets
• figures/: generated figures and plots

Notes

This is an evolving sandbox; scripts and results may change as experiments progress.