CAE-TD3 applies a convolution-attention history encoder inside a TD3 actor-critic to improve continuous control under partial observability, with mixed empirical gains over recurrent baselines.
Outracing champion gran turismo drivers with deep reinforcement learning,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Convolution and Attention Based Encoder for Reinforcement Learning under Partial Observability
CAE-TD3 applies a convolution-attention history encoder inside a TD3 actor-critic to improve continuous control under partial observability, with mixed empirical gains over recurrent baselines.