SCaLE: Switching Cost aware Learning and Exploration
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911374799536128 |
|---|---|
| author | Bhuyan, Neelkamal Mukherjee, Debankur Wierman, Adam |
| author_facet | Bhuyan, Neelkamal Mukherjee, Debankur Wierman, Adam |
| contents | This work addresses the fundamental problem of unbounded metric movement costs in bandit online convex optimization, by considering high-dimensional dynamic quadratic hitting costs and $\ell_2$-norm switching costs in a noisy bandit feedback model. For a general class of stochastic environments, we provide the first algorithm SCaLE that provably achieves a distribution-agnostic sub-linear dynamic regret, without the knowledge of hitting cost structure. En-route, we present a novel spectral regret analysis that separately quantifies eigenvalue-error driven regret and eigenbasis-perturbation driven regret. Extensive numerical experiments, against online-learning baselines, corroborate our claims, and highlight statistical consistency of our algorithm. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_09042 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SCaLE: Switching Cost aware Learning and Exploration Bhuyan, Neelkamal Mukherjee, Debankur Wierman, Adam Machine Learning Data Structures and Algorithms Optimization and Control Probability This work addresses the fundamental problem of unbounded metric movement costs in bandit online convex optimization, by considering high-dimensional dynamic quadratic hitting costs and $\ell_2$-norm switching costs in a noisy bandit feedback model. For a general class of stochastic environments, we provide the first algorithm SCaLE that provably achieves a distribution-agnostic sub-linear dynamic regret, without the knowledge of hitting cost structure. En-route, we present a novel spectral regret analysis that separately quantifies eigenvalue-error driven regret and eigenbasis-perturbation driven regret. Extensive numerical experiments, against online-learning baselines, corroborate our claims, and highlight statistical consistency of our algorithm. |
| title | SCaLE: Switching Cost aware Learning and Exploration |
| topic | Machine Learning Data Structures and Algorithms Optimization and Control Probability |
| url | https://arxiv.org/abs/2601.09042 |