An entropy generation formula on \(RCD(K,\infty)\) spaces (Q1783741)

This paper deals with the existence of the minimal characteristic for the stochastic value function. This is defined as the infimum of a time-depending performance index function over an appropriate set of vector fields (drift) and their corresponding weak solution to the Fokker-Planck equation. The author generalizes the results of \textit{J. Feng} and \textit{T. van Nguyen} [J. Math. Pures Appl. (9) 97, No. 4, 318--390 (2012; Zbl 1242.49059)] by considering a class of metric measure spaces which satisfy the \(RCD(K,\infty)\) property. By using the properties of \(RCD(K,\infty)\) spaces and introducing an appropriate class of test functions, the notion of weak solution for the Fokker-Planck equation is given. Further it is shown that this solution is sufficiently regular. Then, by approximating the drift with the regularized drifts, an entropy generation formula for the weak solution is given. Based on this formula, the existence of minimizers for the value functions is proven. This setting covers also the Fokker-Planck equations defined on bounded domains.