Machine Learning Approaches to Automated Mechanism Design for Public Project Problem

Abstract

Mechanism design is a central research branch in microeconomics. An effective mechanism can significantly improve performance and efficiency of social decisions under desired objectives, such as to maximize social welfare or to maximize revenue for agents.However, mechanism design is challenging for many common models including the public project problem model which we study in this thesis. A typical public project problem is a group of agents crowdfunding a public project (e.g., building a bridge).The mechanism will decide the payment and allocation for each agent (e.g., how much the agent pays, and whether the agent can use it) according to their valuations. The mechanism can be applied to various economic scenarios, including those related to cyber security. There are different constraints and optimized objectives for different public project scenarios (sub-problems), making it unrealistic to design a universal mechanism that fits all scenarios, and designing mechanisms for different settings manually is a taxing job. Therefore, we explore automated mechanism design (AMD) (Sandholm, 2003) of public project problems under different constraints.In this thesis, we focus on the public project problem, which includes many subproblems (excludable/non-excludable, divisible/indivisible, binary/non-binary). We study the classical public project model and extend this model to other related areas such as the zero-day exploit markets. For different sub-problems of the public project problem, we adopt different novel machine learning techniques to design optimal or near-optimal mechanisms via automated mechanism design. xii We evaluate our mechanisms by theoretical analysis or experimentally comparing our mechanisms against existing mechanisms. The experiments and theoretical results show that our mechanisms are better than state-of-the-art automated or manual mechanisms. xiii List of Abbreviations AMD Automated mechanism design GAN Generative adversarial network with him in these three years. I very much appreciate his help. When I stock in no research inspiration, he can always figure it out and find a suitable way for me to solve the problems. Mingyu is a humble scholar. His research foresight and research inspiration is lifelong treasure for me. I would thank Dr. Wei Zhang, and professor Muhammad Ali Babar. They provide some suggestions and comments that helped me to significantly improve the content of this thesis and papers. And from my Ph.D. role, I also thank my collaborators, lab meta, and my friends for the help with the papers. I would like to mention Ba-Dung Le, Runqi Guo, Yuko Sakurai, Wuli Zuo, Xie Yue, Zhigang Lu for many useful discussions. My study area is not a hot area in this university. Thanks, you give me many useful discussions. Good luck to you in the future. And I will thank Wuli Zuo for checking and correcting my grammar mistakes. I also want to thank my parents and my grandparents for financially supporting me during my Ph.D. study. And because of Covid-19, it is a hard time to connect you only o...