Stochastic optimisation and control applied to finance (Q2382317): Difference between revisions

The book is an original presentation of classical and recent techniques for stochastic control problems and their applications in mathematical finance. Even if it does not always cover the most general context it provides a rather quick path to advanced results in the field, still having the goal of a friendly presentation starting from intuitive arguments and proceeding to rigorous statements and (sometimes original) proofs. I consider this book a very useful tool for students and researchers who want to reach rapidly an adequate knowledge of modern techniques in stochastic control to be able to enter the recent literature concerning the applications of stochastic control methods to mathematical finance. In fact, recent papers and books in the field are of advanced technical level and rather difficult for nonspecialists so that the book fills a gap providing a unified presentation of technical methods and their applications to finance. Also, the initial chapter of the book is an effective survey of basic results in stochastic analysis so that the overall presentation is self-contained and particularly useful to students and researchers who were only marginally exposed to the methods and techniques of stochastic analysis and control. The organization of the book will now be outlined. Chapter 2 deals with the structure of stochastic control problems and their applications in finance. Such problems will be studied in detail in subsequent chapters and solved using the classical and modern control techniques developed in the book. Chapter 3 studies classical dynamic programming techniques for the solution to stochastic control problems. A formal proof of Bellman Dynamic Programming Principle (DPP) is provided as well as the formal derivation of its infinitesimal counterpart, namely the Hamilton-Jacobi-Bellman (HJB) equation is given. The classical approach to stochastic control problems, based on the verification theorem, is developed and a few applications in finance (Merton's finite horizon portfolio selection problem and an infinite horizon production and consumption problem) are discussed. The limitations of the method are due to the requirement of regularity of the value function (representing the optimal gain- or cost-to-go) and are illustrated on examples where the value function is not regular (singular control problems). Chapter 4 deals with the approach to dynamic programming based on viscosity solutions to the HJB equation. This chapter provides a basic introduction to the notion of viscosity solutions and to the results useful for their applications in stochastic control. Indeed, viscosity solutions provide a rigorous formulation of the HJB equation for locally bounded value functions. This is due to the fact that, using comparison results for viscosity solutions it can be shown under mild assumptions that the value function is the unique solution to the HJB variational inequality associated to the DDP. Examples not covered by the classical approach, namely problems of irreversible investment for a firm and super-replication prices with uncertain volatility, conclude the chapter. Chapter 5 is concerned with the study of Backward Stochastic Differential Equations (BSDE) and their applications in stochastic control and in particular for the solution to problems in mathematical finance. The precise notion of BSDE is discussed and a basic introduction to the theory of BSDE in view of its applications in finance is provided. Existence and uniqueness of solutions and a comparison principle as well as representation of PDE by solutions to BSDE are discussed. Then it is shown that BSDE allow a generalization of the stochastic maximum principle and its sufficiency for stochastic optimal control is proved. Applications in finance include the maximization of exponential utility of a contingent claim and portfolio optimization with mean-variance criterion. The final Chapter 6 deals with the solution to stochastic control problems using martingale methods and convex duality. The general idea is to transform the original problem to an equivalent one where the maximization is with respect to the state variable and with a linear constraint given in terms of a dual variable. This dual problem can be solved using convex analysis techniques. The main tool of this method is a recent result in stochastic analysis called optional decomposition of supermartingales. Problems of portfolio optimization and mean-variance hedging are finally discussed.