March 14th Colloquium: AI for Social Impact: Learning and Optimization in Network and Combinatorial Spaces -- Haipeng Chen

Talk Abstract: Decision-making has been ubiquitous in socially important domains such as public health, cybersecurity, and social networks. Many of the decision-making problems essentially have a network or combinatorial structure, which are known to be challenging to solve because of the exponentially large solution spaces. The challenge has been further amplified by real-world characteristics of the societal problems, such as stochasticity, sequential decision-making, and unknown constraints/objectives. This talk covers two lines of my research towards solving these decision-making problems using machine learning methods. In the first line of works that are motivated by public health problems such as epidemics control and network-based preventative health, I will introduce how I design novel reinforcement learning methods for sequential network intervention under stochasticity. In the second line of works that are motivated by the cybersecurity problem of dynamic vulnerability patching, I will talk about how prediction and reinforcement learning are combined to solve sequential stochastic combinatorial optimization problems where the model parameters are unknown.

Location: 5317 Sennott Square, 10:00-11:00 a.m. on Monday, March 14th, 2022.

Biosketch: Haipeng Chen is a CRCS postdoctoral fellow at Harvard University. He completed his PhD from Nanyang Technological University. His primary research interest lies in AI for social impact. For AI techniques, he focuses on reinforcement learning, prediction, and optimization. For social domains, he is interested in public health, cybersecurity, and transportation, particularly in problems with a network structure. His research has been recognized with the best paper nomination at AAMAS-2021, Innovation Demonstration Award runner-up at IJCAI-2019, and Champion of the 2017 Microsoft Malmo Collaborative AI Challenge. He has published in premier AI conferences (e.g., AAAI, IJCAI, NeurIPS, AAMAS, UAI, KDD, and ICDM), and journals (e.g., IEEE/ACM Transactions). His work has been covered by popular press such as The Wall Street Journal, Scientific American, Digital Guardian, ScienceBlog, and AAAS. As part of his research agenda, he has established partnerships with non-profits such as The Family Van, Mobile Health Map, and Safe Place for Youth.

View this event on the School of Computing and Information events calendar.

News Type