Asymptotically Optimal Problem-Dependent Bandit Policies for Transfer Learning
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916964684791808 |
|---|---|
| author | Prevost, Adrien Mathieu, Timothee Maillard, Odalric-Ambrym |
| author_facet | Prevost, Adrien Mathieu, Timothee Maillard, Odalric-Ambrym |
| contents | We study the non-contextual multi-armed bandit problem in a transfer learning setting: before any pulls, the learner is given N'_k i.i.d. samples from each source distribution nu'_k, and the true target distributions nu_k lie within a known distance bound d_k(nu_k, nu'_k) <= L_k. In this framework, we first derive a problem-dependent asymptotic lower bound on cumulative regret that extends the classical Lai-Robbins result to incorporate the transfer parameters (d_k, L_k, N'_k). We then propose KL-UCB-Transfer, a simple index policy that matches this new bound in the Gaussian case. Finally, we validate our approach via simulations, showing that KL-UCB-Transfer significantly outperforms the no-prior baseline when source and target distributions are sufficiently close. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_19098 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Asymptotically Optimal Problem-Dependent Bandit Policies for Transfer Learning Prevost, Adrien Mathieu, Timothee Maillard, Odalric-Ambrym Machine Learning Statistics Theory We study the non-contextual multi-armed bandit problem in a transfer learning setting: before any pulls, the learner is given N'_k i.i.d. samples from each source distribution nu'_k, and the true target distributions nu_k lie within a known distance bound d_k(nu_k, nu'_k) <= L_k. In this framework, we first derive a problem-dependent asymptotic lower bound on cumulative regret that extends the classical Lai-Robbins result to incorporate the transfer parameters (d_k, L_k, N'_k). We then propose KL-UCB-Transfer, a simple index policy that matches this new bound in the Gaussian case. Finally, we validate our approach via simulations, showing that KL-UCB-Transfer significantly outperforms the no-prior baseline when source and target distributions are sufficiently close. |
| title | Asymptotically Optimal Problem-Dependent Bandit Policies for Transfer Learning |
| topic | Machine Learning Statistics Theory |
| url | https://arxiv.org/abs/2509.19098 |