Existence of solutions in non-convex dynamic programming and optimal investment (Q513744)

The authors of this article analyze the main provisions of the existence of solutions in non-convex dynamic programming and optimal investments. Since there is a growing interest in unarmed scraps, the topic of the article is relevant at the moment. Earlier, many authors considered the so-called ``S-shaped'' utilities, which are convex (from the point of view of risk) to a certain level of wealth and concave (not prone to risk) over it. The authors overcome this lack of concavity by applying new mathematical tools. In my opinion, the main result of this paper is the proof of a fairly general dynamic programming principle for discrete time problems with multistage stochastic optimization with a not necessarily concave objective function described in Theorem 3. The authors adopt a general stochastic dynamic programming format that extends the more well-known problems of stochastic control in discrete time. Also, the results of the existence of previous achievements in the field under consideration in the article are extending, weakening their assumptions about compactness and convexity. The results of the existence of optimal strategies in general are presented. Semimartingale models of markets without friction for concave utility functions are considered. The theorem presented in the article 4 is the result of work based on the proof of the existence of utilities for optimal investment in illiquid markets with discrete time and with bounded top. From the disadvantages of this paper, one can single out the fact that the simple result of this study is of little practical significance, since nothing is known about the solutions of the optimization problem in this field. Also, the model considered in the work does not presuppose the existence of a cash account.