Metraipolis › archive › dc_be82eda0d8eb

Neural Actor--Critic Learning with Minimax Critics and Reflected Langevin Sampling — arXiv:2611.03352

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.03352
License (as archived)
CC0-1.0
Archived
2026-11-03T11:26:47.770Z
SHA-256 of the archived text
3e58b0876eacb7b4c540d2eee8997339e61bd0505973cc1d4ad4728949f23ffd
Stable URL of this page
https://metraipolis.com/documento/dc_be82eda0d8eb

How to cite

Neural Actor--Critic Learning with Minimax Critics and Reflected Langevin Sampling — arXiv:2611.03352 — arXiv (en), https://arxiv.org/abs/2611.03352 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_be82eda0d8eb (Metraipolis archive dc_be82eda0d8eb, sha256 3e58b0876eacb7b4c540d2eee8997339e61bd0505973cc1d4ad4728949f23ffd)
BibTeX
@misc{dcbe82eda0d8eb,
  title = {Neural Actor--Critic Learning with Minimax Critics and Reflected Langevin Sampling — arXiv:2611.03352},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.03352},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_be82eda0d8eb; sha256 3e58b0876eacb7b4c540d2eee8997339e61bd0505973cc1d4ad4728949f23ffd},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.03352. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Actor--critic methods rely on critic gradients to improve continuous-action policies, yet finite-sample guarantees must control both value and gradient errors and their propagation through the actor. We develop a neural actor-critic procedure that combines a regularized minimax Bellman estimator over sigmoid networks with a finite reflected Langevin sampler driven by the fitted critic gradient. Under uniform Bellman regularity and covered reset sampling, a positive high-order action-derivative penalty yields value and action-gradient risk bounds with classical nonparametric minimax exponents up to logarithmic factors. We propagate estimation, mixing, discretization, temperature, and policy-iteration errors to obtain return guarantees, with refinements for bounded multimodal potentials and strongly concave regimes. Empirical studies across controlled simulations, logged continuous-control data, and a simulated decision task show favorable performance relative to the considered baselines.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (stat.ML) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2611.03352 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-03T11:02:37.908Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).