← Search

Kakei Yamamoto

2 accepted papers

2024

Mean Field Langevin Actor-Critic: Faster Convergence and Global Optimality beyond Lazy Learning

ICML 2024poster

This work explores the feature learning capabilities of deep reinforcement learning algorithms in the pursuit of optimal policy determination. We particularly examine an over-parameterized neural actor-critic framework within the mean-field regime, where both actor and critic components undergo upda…

Cited by 1SourcePDFScholar
2024

Symmetric Mean-field Langevin Dynamics for Distributional Minimax Problems

ICLR 2024spotlight

In this paper, we extend mean-field Langevin dynamics to minimax optimization over probability distributions for the first time with symmetric and provably convergent updates. We propose \emph{mean-field Langevin averaged gradient} (MFL-AG), a single-loop algorithm that implements gradient descent a…

Cited by 10SourcePDFScholar