{"id":"arxiv/2604.13259","url":"https://arxiv.org/abs/2604.13259","tags":["math.DS"],"score":0,"title":"Global attractors and fast-slow reduction for finite-state actor-critic mean dynamics","domain":"math.DS","source":"arxiv","authors":["Vladyslav Prytula"],"reviews":0,"subject":"math.DS","summary":"When a learning algorithm reshapes the data distribution it trains on, the long-run behavior depends on the joint evolution of the policy, the value estimate, and the data distribution. We study finite-state actor-critic mean dynamics on the enlarged phase space $(θ,w,μ)$, where $θ$ is the actor parameter, $w$ is an auxiliary critic state, and $μ$ is a state-law variable (the distribution over states induced by the current policy). The state-law coordinate follows the exact controlled-Markov equation $δ\\dotμ= Q_θ^*μ$. Under a softmax actor with box confinement (a smooth proxy for parameter…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-04-14","reactions":{}}