Shinji Ito

Semi-gradient TD learning with neural networks is provably stable and exponentially converges under a simple entropy-induced relaxation condition.

Surrogate probability functions enable FTPL to achieve BOBW guarantees while maintaining computational simplicity, revolutionizing adaptive learning rates in online learning.

Unconstrained bandit linear optimization can be surprisingly reduced to standard online linear optimization using a perturbation approach, unlocking new regret guarantees and high-probability bounds.