🕹️ Deep Q-Network — Space Invaders
Watch trained DQN agents play Atari Space Invaders in real-time. Three variants trained from raw pixels using PyTorch.
Agent Variant
About the agents:
- Baseline DQN — Standard architecture
- Double DQN ⭐ — Reduces Q-value overestimation
- Dueling DQN — Separates state value from action advantage