Pre-specified eigenoptions can speed up credit assignment in tabular gridworlds, but their benefits shrink when options are learned online or approximated with neural networks.
To learn eigenoption policies, we store all samples in a dataset and sweep over the dataset Nsweeps = 100 times using Q-learning with γ = 0.9 and α = 0.1
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 Study of Value-Aware Eigenoptions
Pre-specified eigenoptions can speed up credit assignment in tabular gridworlds, but their benefits shrink when options are learned online or approximated with neural networks.