DQN + Navigation + Lunar variants

Value-based methods and DQN lineage

The strongest bridge from the course tutorials into project work and later experiments.

Good base after curation Mixed modern + legacy

Open Lunar Lander recovery page See archive context

How It Works

Structure

This branch starts with the main DQN tutorial, extends into the archived Value-based Methods project bundle, then reaches the Banana Navigation project and your later LunarLander variants like Double DQN and prioritized replay.

Why Reuse

Role in the new DRL arm

It gives you both the clean reference implementation and the messier follow-on experiments that show how you were exploring beyond the coursework.

Key Assets

What is already here

Core DQN Tutorial

notebook set

Reference DQN notebooks plus `dqn_agent.py` and `model.py`.

source-material/dqn

Navigation Project

project bundle

Banana environment notebooks, report, and reusable DQN project code.

source-material/classwork/project-reports/p1_navigation

Archived Navigation Weights

zip archive

Contains `trained_weights.pth` and `benchmark_weights.pth` inside the archive.

source-material/classwork/project-reports/p1_navigation/p1_project.zip

Banana Environment

unity bundle

Bundled Windows Unity environment for project 1.

source-material/classwork/project-reports/Banana_Windows_x86_64.zip

Lunar DQN Experiments

experiment branch

Double DQN, prioritized replay, and alternate preprocessing ideas.

source-material/lunar/dqn