How park-jihoo/rl_til works
Project Type and Purpose
• This is a multi-course, multi-assignment repository focused on reinforcement learning. It includes implementations (e.g., Wordle RL agents, baby_robot experiments), a set of academic assignments from Stanford (ass1, ass2, ass3, and 2023 variants) and Cornell, plus a library of relevant research papers.
• There isn’t one monolithic application, but rather multiple experiments, environments, and algorithm implementations. The common goal is to demonstrate various RL methods (e.g., deep Q-learning, policy gradients, and variations used in simulated environments) applied to different tasks.
Analyzing the File Structure
• Top-Level Organization:
• “cornell-cs6789”: Contains resources for the Cornell course (likely including its own README) and possibly assignment instructions or code.
• “stanford-cs234”: Contains multiple assignment files and directories (ass1, ass2, ass2_2023, ass3, ass3_2023) with code, configurations, and experimental results.
• “implementation”: Holds independent projects such as “Wordle” (with RL agents implemented in dqn.py, policygradient.py, and human-based approaches) and “baby_robot” which appears to be a separate RL or simulation project.
• “papers”: Consists of markdown files for research papers on bandit algorithms and related RL topics.
• Within each course directory (especially Stanford CS234), the structure is further divided into assignments (e.g., different “ass” folders) that include:
• Configuration files (e.g., Makefiles, YAML files, configs)
• Core RL algorithm implementations (e.g., deep_q_learning, q_learning, ppo.py, policy_gradient.py)
• Utility functions and test environments
• Results directories that capture experiment data, logs, and model weights.