Augmenting each sampled action with its full symmetry orbit makes finite-sample planning exactly equivariant and speeds up learning on several rotationally symmetric control tasks.
Symmetry in Markov decision processes and its implications for single agent and multi agent learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Equivariant Action Sampling for Reinforcement Learning and Planning
Augmenting each sampled action with its full symmetry orbit makes finite-sample planning exactly equivariant and speeds up learning on several rotationally symmetric control tasks.