Metraipolis › archive › dc_053fc46ed690
Correcting Within-Group Self-Selection Bias in Prioritized Replay — arXiv:2609.25297
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.25297
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T15:07:21.981Z
- SHA-256 of the archived text
5040d431f98be8ff27e98969e1c6fa8cdd63d67a177a206a8063be271b7874f3- Stable URL of this page
https://metraipolis.com/documento/dc_053fc46ed690
How to cite
Correcting Within-Group Self-Selection Bias in Prioritized Replay — arXiv:2609.25297 — arXiv (en), https://arxiv.org/abs/2609.25297 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_053fc46ed690 (Metraipolis archive dc_053fc46ed690, sha256 5040d431f98be8ff27e98969e1c6fa8cdd63d67a177a206a8063be271b7874f3)
BibTeX
@misc{dc053fc46ed690,
title = {Correcting Within-Group Self-Selection Bias in Prioritized Replay — arXiv:2609.25297},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.25297},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_053fc46ed690; sha256 5040d431f98be8ff27e98969e1c6fa8cdd63d67a177a206a8063be271b7874f3},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.25297. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Prioritized experience replay (PER) improves sample efficiency by replaying high-priority transitions, usually according to absolute temporal-difference error. In stochastic environments, PER can distort the distribution of realized outcomes replayed from transitions with the same state-action pair. We call this within-group self-selection. We quantify the resulting changes in within-group outcome frequencies and mean Bellman targets. We decompose PER into between-group allocation and conditional sibling selection, and derive fixed-buffer corrections that preserve current group-level priority mass: SAMPLE selects a group through PER and trains on a uniformly sampled sibling; AVG averages sibling Bellman targets; and MODEL samples from an empirical full-outcome model. In exact state-action environments with rare high-magnitude outcomes, sibling-aware replay improves learning efficiency over PER, although matched parameter sweeps show that tuning can narrow some gaps. In MinAtar, approximate VQ-VAE groups with SAMPLE mitigate degradation under mean-preserving reward tails in four of five games. Sibling-aware replay thus retains the focus on high-priority state-action regions while recovering their empirical outcome frequencies. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-23 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.25297 — 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 2026-10-08T14:37:29.383Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-23?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).