Parameter noise for exploration — how is it different?
mediumAnswer
- Plappert et al. 2017: add noise directly to policy parameters (θ + σε) instead of action noise.
- Produces consistent behavior over an episode (θ is fixed until reset) → temporally-correlated exploration.
- Contrast: action noise (ε-greedy, OU noise) is uncorrelated.
- Adaptive σ tuned to match target action-level variability.
- Foundation of noisy nets.
- Empirically better on locomotion.
Check yourself — multiple choice
- Same as action noise
- Add noise to policy weights (θ + σε), fixed for episode → temporally-correlated exploration vs uncorrelated action noise (ε-greedy / OU); foundation of noisy nets; better on locomotion
- Random
- Not real
Parameter noise: noise on θ → temporally correlated exploration.
#exploration
Practise Reinforcement Learning
214 interview questions in this topic.