A policy Newton method for RKHS-based reinforcement learning is derived, but the claimed finite-dimensional equivalence wrongly drops the Gram matrix from the cubic regularization term.
Super-universal regularized newton method
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Policy Newton Algorithm in Reproducing Kernel Hilbert Space
A policy Newton method for RKHS-based reinforcement learning is derived, but the claimed finite-dimensional equivalence wrongly drops the Gram matrix from the cubic regularization term.