QV-learning has an expected contraction for on-policy prediction, QVMAX is biased for off-policy control, and a new regularized dueling algorithm (RDQ) outperforms Dueling DQN on MinAtar.
Bertsekas and John N
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
-
An Analysis of Action-Value Temporal-Difference Methods That Learn State Values
QV-learning has an expected contraction for on-policy prediction, QVMAX is biased for off-policy control, and a new regularized dueling algorithm (RDQ) outperforms Dueling DQN on MinAtar.