ε-greedy — how do you schedule ε?
easyAnswer
- Start ε high (0.5-1.0) for broad exploration, anneal linearly or exponentially to a small floor (0.01-0.1).
- Atari DQN: linear 1.0 → 0.1 over first 1M steps, then 0.1 → 0.01 over 24M more.
- Rule: schedule matches expected training length.
- Alternative decays: 1/n schedule (asymptotic optimality for tabular bandits), exponential.
- Modern deep RL often uses noisy nets or entropy bonus instead of ε-greedy.
Check yourself — multiple choice
- Fixed ε
- Anneal ε from ~1.0 → 0.01-0.1 (linear or exp); DQN-Atari: 1.0→0.1 over 1M, then to 0.01; length-matched schedule; noisy nets / entropy bonus in modern deep RL
- Random
- Only 0.5
ε schedule: anneal from ~1 to 0.01-0.1; length matched to training.
#exploration
Practise Reinforcement Learning
214 interview questions in this topic.