Choices and intervals (Q2630873)

\textit{H. J. Kushner} and \textit{D. S. Clark} [Stochastic approximation methods for constrained an unconstrained systems. Applied Mathematical Sciences 26. New York-Heidelberg-Berlin: Springer-Verlag (1978; Zbl 0381.60004)] developed general methods for convergence results, in finite-dimensional settings, for stochastic approximation. The authors extend their methods to an infinite dimensional setting. More precisely, the authors consider a random interval splitting process, where the splitting rule depends on the empirical distribution of interval lengths. It is shown that the empirical distribution converges to a limit almost surely as the number of intervals goes to infinity. The resulting limit is characterized by an ODE, which is used to provide the tail estimates of the limit distribution. Moreover, the convergence is established by showing that the size-based empirical distribution evolves in the limit according to a certain deterministic evolution equation.