Reinforcement Learning and Stochastic Optimization

Acknowledgments

The foundation of this book is a modeling framework for sequential decision problems that involves searching over four classes of policies for making decisions. The recognition that we needed all four classes of policies came from working on a wide range of problems spanning freight transportation (almost all modes), energy, health, e-commerce, finance, and even materials science (!!).

This research required a lot of computational work, which was only possible through the efforts of the many students and staff that worked in CASTLE Lab. Over my 39 years of teaching at Princeton, I benefited tremendously from the interactions with 70 graduate students and post-doctoral associates, along with nine professional staff. I am deeply indebted to the contributions of this exceptionally talented group of men and women who allowed me to participate in the challenges of getting computational methods to work on such a wide range of problems. It was precisely this diversity of problem settings that led me to appreciate the motivation for the different methods for solving problems. In the process, I met people from across the jungle, and learned to speak their language not just by reading papers, but by talking to them and, often, working on their problems.

I would also like to acknowledge what I learned from supervising over 200 senior theses. While not as advanced as the graduate research, the undergraduates helped expose me to an even wider range of problems, spanning topics ...