Stochastic Switching for Partially Observable Dynamics and Optimal Asset Allocation

In industrial appli ations, optimal c ontrol problems frequently appear in the c ontext of dec isions-making under inc omplete information. In su ch framework, dec isions must be adapted dynami cally to acc ount for possible regime c hanges of the underlying dynamic s. Using sto hastic filtering theory, Markovian evolution c an be modeled in terms of latent variables, whi h naturally leads to high dimensional state spa ce, making prac tic al solutions to these c ontrol problems notoriously c hallenging. In our approa h, we utilize a speci fic stru ture of this problem c lass to present a solution in terms of simple, reliable, and fast algorithms.