Eluder dimension: localise it!
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918456076533760 |
|---|---|
| author | Bakhtiari, Alireza Ayoub, Alex Robertson, Samuel Janz, David Szepesvári, Csaba |
| author_facet | Bakhtiari, Alireza Ayoub, Alex Robertson, Samuel Janz, David Szepesvári, Csaba |
| contents | We establish a lower bound on the eluder dimension of generalised linear model classes, showing that standard eluder dimension-based analysis cannot lead to first-order regret bounds. To address this, we introduce a localisation method for the eluder dimension; our analysis immediately recovers and improves on classic results for Bernoulli bandits, and allows for the first genuine first-order bounds for finite-horizon reinforcement learning tasks with bounded cumulative returns. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_09825 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Eluder dimension: localise it! Bakhtiari, Alireza Ayoub, Alex Robertson, Samuel Janz, David Szepesvári, Csaba Machine Learning We establish a lower bound on the eluder dimension of generalised linear model classes, showing that standard eluder dimension-based analysis cannot lead to first-order regret bounds. To address this, we introduce a localisation method for the eluder dimension; our analysis immediately recovers and improves on classic results for Bernoulli bandits, and allows for the first genuine first-order bounds for finite-horizon reinforcement learning tasks with bounded cumulative returns. |
| title | Eluder dimension: localise it! |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2601.09825 |