Stochastic variational formulas for solutions to linear diffusion equations
Joseph G. Conlon
University of Michigan, Ann Arbor, USAMohar G. Guha
University of Michigan, Ann Arbor, USA
Abstract
This paper is concerned with solutions to a one-dimensional linear diffusion equation and their relation to some problems in stochastic control theory. A stochastic variational formula is obtained for the logarithm of the solution to the diffusion equation, with terminal data which is the characteristic function of a set. In this case the terminal data for the control problem is singular, and hence standard theory does not apply. The variational formula is used to prove convergence, in the zero noise limit, of the cost function for the stochastic control problem and its first derivatives, to the corresponding quantities for a classical control problem.
Cite this article
Joseph G. Conlon, Mohar G. Guha, Stochastic variational formulas for solutions to linear diffusion equations. Rev. Mat. Iberoam. 30 (2014), no. 2, pp. 581–666
DOI 10.4171/RMI/794