Vikas Garg: Generalization and Representational Limits of Graph Neural Networks
Speaker: Vikas Garg (Aalto University)
Title: Generalization and Representational Limits of Graph Neural Networks.
Abstract: Graphs provide a natural abstraction to model relational and strategic data in domains as diverse as biology (e.g., molecules), multiagent settings (e.g., online vendors on ecommerce platforms), and distributed systems (e.g., Internet). Graphs also find much use as theoretical objects (e.g., probabilistic graphical models), and several important algorithms (e.g., max-flow for image segmentation) can be invoked when tasks are formulated in terms of graphs.
Bio: Vikas Garg is an Assistant Professor at Aalto University. He recently graduated from MIT, where his PhD in Computer Science was supervised by Prof. Tommi Jaakkola. His research interests in machine learning include generative models, graphical models, theory of deep learning, and learning under uncertainty or resource constraints along with their intersections with optimization and game theory.