Generic

Recent researches in graph learning not only demonstrated its power in solving tasks that naturally involve graphs / networks as inputs, but also shown promises in using graph modeling and learning for tasks that were not traditionally thought of as graphs, such as autoML, natural languages, database ML, physical simulations etc. Our goal is to push the boundary of graph deep learning and leverage the power of relational structure in solving the most challenging tasks.

To learn effective representations for structured data, we develop effective graph neural network (GNN) architectures. we further leverage a vareity of geometric approaches, such as order, box, hyperbolic and cone embeddings. The properties inherent in the geometric structures allow us to achieve superior expressive power and generalization when modeling the inductive biases of real-world data.

Graph Neural Networks

Donec eget ex magna. Interdum et malesuada fames ac ante ipsum primis in faucibus. Pellentesque venenatis dolor imperdiet dolor mattis sagittis. Praesent rutrum sem diam, vitae egestas enim auctor sit amet. Pellentesque leo mauris, consectetur id ipsum sit amet, fergiat. Pellentesque in mi eu massa lacinia malesuada et a elit. Donec urna ex, lacinia in purus ac, pretium pulvinar mauris. Curabitur sapien risus, commodo eget turpis at, elementum convallis elit. Pellentesque enim turpis, hendrerit.

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis dapibus rutrum facilisis. Class aptent taciti sociosqu ad litora torquent per conubia nostra, per inceptos himenaeos. Etiam tristique libero eu nibh porttitor fermentum. Nullam venenatis erat id vehicula viverra. Nunc ultrices eros ut ultricies condimentum. Mauris risus lacus, blandit sit amet venenatis non, bibendum vitae dolor. Nunc lorem mauris, fringilla in aliquam at, euismod in lectus. Pellentesque habitant morbi tristique senectus et netus et malesuada fames ac turpis egestas. In non lorem sit amet elit placerat maximus. Pellentesque aliquam maximus risus, vel sed vehicula. Interdum et malesuada fames ac ante ipsum primis in faucibus. Pellentesque venenatis dolor imperdiet dolor mattis sagittis. Praesent rutrum sem diam, vitae egestas enim auctor sit amet. Pellentesque leo mauris, consectetur id ipsum sit amet, fersapien risus, commodo eget turpis at, elementum convallis elit. Pellentesque enim turpis, hendrerit tristique lorem ipsum dolor.

Geometric Representation Learning

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Duis dapibus rutrum facilisis. Class aptent taciti sociosqu ad litora torquent per conubia nostra, per inceptos himenaeos. Etiam tristique libero eu nibh porttitor fermentum. Nullam venenatis erat id vehicula viverra. Nunc ultrices eros ut ultricies condimentum. Mauris risus lacus, blandit sit amet venenatis non, bibendum vitae dolor. Nunc lorem mauris, fringilla in aliquam at, euismod in lectus. Pellentesque habitant morbi tristique senectus et netus et malesuada fames ac turpis egestas. In non lorem sit amet elit placerat maximus. Pellentesque aliquam maximus risus, vel sed vehicula.

Interdum et malesuada fames ac ante ipsum primis in faucibus. Pellentesque venenatis dolor imperdiet dolor mattis sagittis. Praesent rutrum sem diam, vitae egestas enim auctor sit amet. Pellentesque leo mauris, consectetur id ipsum sit amet, fersapien risus, commodo eget turpis at, elementum convallis elit. Pellentesque enim turpis, hendrerit tristique lorem ipsum dolor.