Professor Xia Li of Our School Published Paper as Sole Author in the UTD24 Journal Production and Operations Management

Recently, the Production and Operations Management (POM), an international high-level journal in the field of management, published on-line the important research result Risk-Sensitive Markov Decision Processes with Combined Metrics of Mean and Variance by Professor Xia Li of our school as sole author. This is an important manifestation of the high-quality research results of our school and contributes to the "double first-class" discipline construction of our school.

This research focuses on the practical problems of portfolio management optimization. The variance of asset returns reflects the volatility of financial assets, and is an important indicator to measure the risk of financial assets. Mean-variance optimization is a classic problem in portfolio management optimization, whose goal is not only to maximize the average return on assets, but also control the variance of assets return. Mean-variance optimization is a representative result of Nobel Prize winner Professor Markowitz. The early mean-variance optimization is only for static problem scenarios. When extended to dynamic system, the problem often needs to be modeled as a Markov Decision Process. However, due to the time inconsistency of the variance index, the classic dynamic programming principle no longer holds, and the tradition optimization theory cannot directly solves the problem. This research solves the problem of joint optimization of mean and variance in a random dynamic environment, gives the optimization algorithm for the optimal strategy, analyzes and proves the convergence properties of the algorithm, and is expected to achieve further research results on the international frontier subject of risk-sensitive reinforcement learning.

Since 2014, Professor Xia Li has continuously conducted in-depth research on this problem, studied the problem from a new perspective of sensitivity optimization theory, and proposed a new method to solve the random dynamic mean-variance optimization problem, which has been recognized by international peers. In the past five years, he has published a number of papers as sole author on this issue, including top journal in management POM (1 article), top journal in control Automatica (2 articles), and authoritative journal in applied mathematics DEDS (1 article). He obtained a general project of the National Natural Science Foundation of China (Risk-Sensitive Markov Decision Processes and Reinforcement Learning and its Application, 62073346) and a number of invention patents, and guided students and post-doctoral fellows to further study the implementation of risk-sensitive reinforcement learning algorithm based on this methodology, which has been applied to financial portfolio management optimization, suppression of volatility of new energy power generation, and other engineering problems, forming a relatively complete research system.