DIAYN — Diversity is All You Need for skill discovery.
hardAnswer
- Eysenbach et al. 2018.
- Unsupervised skill learning: (1) sample skill z from prior, (2) train to maximize mutual information between z and visited states s: max I(s; z).
- Uses a discriminator trained to identify skill from state.
- Yields diverse skills without external reward.
- Foundation of unsupervised RL and skill libraries for downstream tasks.
Check yourself — multiple choice
- Random
- Sample skill z from prior, train to maximize MI(s; z) via discriminator ; yields diverse skills without external reward; unsupervised RL foundation
- Same as BC
- Not real
DIAYN: max MI(s; z) via discriminator → diverse skills without reward.
#exploration#policy-methods
Practise Reinforcement Learning
214 interview questions in this topic.