from    
to    
search  

 


Symmetry restoration and quantum Mpemba effects in chaotic andlocalization sy...
Quantum Gases 2024
Stories of Fermions in an Optical Box
Contractive Unitary and Classical Shadow Tomography
报告题目:
AI-driven Mechanism Design
 报告人:
Weiran Shen
postdoc researcher at Carnegie Mellon University
报告时间:
2019-12-23 14:00
报告地点:
电子工程馆(罗姆楼)5层206会议室
主办单位:
电子工程系
  简介:

ABSTRACT:

Mechanism design has become a central research topic at the  interface of computer science and economics. Despite decades of efforts, there  is still a huge gap between its theory and application. The strong assumptions  made by the theory usually do not hold in practice, and the huge design space in  real-world applications poses serious challenges to theoretical  analysis.

To bridge the gap, we propose the AI-driven mechanism design  framework. Our framework contains both an agent model and a mechanism model,  where the agent model describes the agents’ actual behaviors and the mechanism  model represents a parameterized mechanism in a large mechanism space. Our  framework uses AI techniques to model complicated agent behaviors and search for  a mechanism with desirable performances.

The AI-driven mechanism design framework provides a new methodology  to look at mechanism design problems through the lens of AI. We show that this  framework can solve both theoretic and application problems, and that it is able  to tackle challenges in different aspects of mechanism  design.

 

BIOGRAPHY:

Weiran Shen is a postdoc researcher at Carnegie Mellon University.  He obtained his Ph.D. at IIIS, Tsinghua University in 2019. Prior to that, he  obtained his B.E. from the Department of Electronic Engineering, Tsinghua  University. His research interest includes mechanism design, game theory,  multi-agent system, and machine learning. He proposed the reinforcement  mechanism design framework, which has already been adopted by Baidu, and was  highlighted in its 2018 Q1 Financial Report. 


今日相关信息
Robustly Optimal Operation for Activ...
Perspectives of Deep Learning for Pro...
International Journal of Electrical P...
 
同类别相关信息
系列交叉论坛第11期:电磁兼容与电磁安...
博士创“芯”说第二期:动态忆阻器与储备...
学堂班系列讲座:分子聚集体的电子激发态...
解锁深度视频理解的潜力
清华信息大讲堂第192讲:An evolution...
学术活动