In this work, we study the online convex optimization problem with curved losses and delayed feedback. When losses are strongly convex, existing approaches obtain regret bounds of order $d_{\max} \ln T$, where $d_{\max}$ is the maximum delay and $T$ is the time horizon. However, in many cases, this guarantee can be much worse than $\sqrt{d_{\mathrm{tot}}}$ as obtained by a delayed version of online gradient descent, where $d_{\mathrm{tot}}$ is the total delay. We bridge this gap by proposing a variant of follow-the-regularized-leader that obtains regret of order $\min\{\sigma_{\max}\ln T, \sqrt{d_{\mathrm{tot}}}\}$, where $\sigma_{\max}$ is the maximum number of missing observations. We then consider exp-concave losses and extend the Online Newton Step algorithm to handle delays with an adaptive learning rate tuning, achieving regret $\min\{d_{\max} n\ln T, \sqrt{d_{\mathrm{tot}}}\}$ where $n$ is the dimension. To our knowledge, this is the first algorithm to achieve such a regret bound for exp-concave losses. We further consider the problem of unconstrained online linear regression and achieve a similar guarantee by designing a variant of the Vovk-Azoury-Warmuth forecaster with a clipping trick. Finally, we implement our algorithms and conduct experiments under various types of delay and losses, showing an improved performance over existing methods.

Exploiting Curvature in Online Convex Optimization with Delayed Feedback / H. Qiu, E. Esposito, M. Zhang (PROCEEDINGS OF MACHINE LEARNING RESEARCH). - In: International Conference on Machine Learning / [a cura di] A. Singh, M. Fazel, D. Hsu, S. Lacoste-Julien, F. Berkenkamp, T. Maharaj, K. Wagstaff, J. Zhu. - [s.l] : PMLR, 2025. - pp. 50448-50479 (( 42. International Conference on Machine Learning Vancouver 2025.

Exploiting Curvature in Online Convex Optimization with Delayed Feedback

H. Qiu
Co-primo
;
E. Esposito
Co-primo
;
2025

Abstract

In this work, we study the online convex optimization problem with curved losses and delayed feedback. When losses are strongly convex, existing approaches obtain regret bounds of order $d_{\max} \ln T$, where $d_{\max}$ is the maximum delay and $T$ is the time horizon. However, in many cases, this guarantee can be much worse than $\sqrt{d_{\mathrm{tot}}}$ as obtained by a delayed version of online gradient descent, where $d_{\mathrm{tot}}$ is the total delay. We bridge this gap by proposing a variant of follow-the-regularized-leader that obtains regret of order $\min\{\sigma_{\max}\ln T, \sqrt{d_{\mathrm{tot}}}\}$, where $\sigma_{\max}$ is the maximum number of missing observations. We then consider exp-concave losses and extend the Online Newton Step algorithm to handle delays with an adaptive learning rate tuning, achieving regret $\min\{d_{\max} n\ln T, \sqrt{d_{\mathrm{tot}}}\}$ where $n$ is the dimension. To our knowledge, this is the first algorithm to achieve such a regret bound for exp-concave losses. We further consider the problem of unconstrained online linear regression and achieve a similar guarantee by designing a variant of the Vovk-Azoury-Warmuth forecaster with a clipping trick. Finally, we implement our algorithms and conduct experiments under various types of delay and losses, showing an improved performance over existing methods.
online learning; delayed feedback; curved losses
Settore INFO-01/A - Informatica
   European Lighthouse of AI for Sustainability (ELIAS)
   ELIAS
   EUROPEAN COMMISSION
   101120237

   One Health Action Hub: task force di Ateneo per la resilienza di ecosistemi territoriali (1H_Hub) - ONE HEALTH ACTION HUB
   (1H_Hub) - ONE HEALTH ACTION HUB
   UNIVERSITA' DEGLI STUDI DI MILANO
2025
https://proceedings.mlr.press/v267/qiu25a.html
Book Part (author)
File in questo prodotto:
File Dimensione Formato  
qiu25a.pdf

accesso aperto

Tipologia: Publisher's version/PDF
Licenza: Creative commons
Dimensione 8.53 MB
Formato Adobe PDF
8.53 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2434/1208315
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
  • OpenAlex ND
social impact