RIKEN Center for Advanced Intelligence Project Sequential Decision Making Team

Team leader: Shinji Ito (Ph.D.)

Japanese Page

Research Summary

The Sequential Decision Making Team works to develop algorithms and theories for making rational decisions in a sequential manner in the face of forecast uncertainty and environmental fluctuations. In recent years, along with the evolution of information technology, there has been a demand for technology to make rational decisions based on the large amount of data being generated in real-time in today's world. To meet this challenge, we promote research related to online learning, bandit problems, and reinforcement learning, aiming to understand effective decision-making algorithms in a fluctuating environment and to construct and extend theoretical systems that support such algorithms.

Main Research Fields

Informatics

Related Research Fields

Engineering
Mathematical & Physical Sciences
Theory of informatics
Mathematical informatics
Intelligent informatics

Keywords

Sequential decision-making
Online learning
Bandit problems
Reinforcement learning
Learning theory

Selected Publications

1. S. Ito and K. Takemura:
"An Exploration-by-Optimization Approach to Best of Both Worlds in Linear Bandits"
Advances in Neural Information and Processing Systems 36 (NeurIPS), to appear (2023).
2. S. Ito, D. Hatano, H. Sumita, K. Takemura, T. Fukunaga, N. Kakimura, and K.-I. Kawarabayashi:
"Bandit Task Assignment with Unknown Processing Time“
Advances in Neural Information and Processing Systems 36 (NeurIPS), to appear (2023).
3. T. Tsuchiya, S. Ito, and J. Honda:
"Stability-penalty-adaptive follow-the-regularized-leader: Sparsity, game-dependency, and best-of-both-worlds“
Advances in Neural Information and Processing Systems 36 (NeurIPS), to appear (2023).
4. S. Ito and K. Takemura:
"Best-of-Three-Worlds Linear Bandit Algorithm with Variance-Adaptive Regret Bounds"
Proceedings of 36th Conference on Learning Theory (COLT), pp. 2653-2677 (2023).
5. T. Tsuchiya, S. Ito, and J. Honda:
"Further Adaptive Best-of-Both-Worlds Algorithm for Combinatorial Semi-Bandits"
Proceedings of The 26th International Conference on Artificial Intelligence and Statistics (AISTATS), pp. 8117-8144 (2023).
6. J. Honda, S. Ito, and T. Tsuchiya:
"Follow-the-Perturbed-Leader Achieves Best-of-Both-Worlds for Bandit Problems"
Proceedings of The 34th International Conference on Algorithmic Learning Theory (ALT), pp. 726-754 (2023).
7. T. Tsuchiya, S. Ito, and J. Honda:
"Best-of-Both-Worlds Algorithms for Partial Monitoring"
Proceedings of The 34th International Conference on Algorithmic Learning Theory (ALT), pp. 1484-1515 (2023).
8. S. Ito, T. Tsuchiya, and J. Honda:
"Nearly Optimal Best-of-Both-Worlds Algorithms for Online Learning with Feedback Graphs"
Advances in Neural Information and Processing Systems 35 (NeurIPS), pp. 28631-28643 (2022).
9. S. Ito:
"Revisiting Online Submodular Minimization: Gap-Dependent Regret Bounds, Best of Both Worlds and Adversarial Robustness"
Proceedings of the 39th International Conference on Machine Learning (ICML), pp. 9678-9694 (2022).
10. S. Ito, T. Tsuchiya, and J. Honda:
"Adversarially Robust Multi-Armed Bandit Algorithm with Variance-Dependent Regret Bounds"
Proceedings of 36th Conference on Learning Theory (COLT), pp. 1421-1422 (2022).

Lab Members

Principal investigator

Shinji Ito: Team leader

Core members

Junya Honda: Visiting Scientist
Taira Tsuchiya: Visiting Scientist
Junpei Komiyama: Visiting Scientist
Kazushi Tsutsui: Visiting Scientist

Contact Information

Nihonbashi 1-chome Mitsui Building, 15th floor,
1-4-1 Nihonbashi, Chuo-ku, Tokyo
103-0027, Japan
Email:shinji.ito.hh [at] riken.jp