Like Hui is a PhD student from UC San Diego, working with Misha Belkin. Her research interest is on the fundamental understanding of generalization and optimization in machine learning. She did some empirical work trying to reduce the gap between practice and theory, including a comparison of the square loss and the cross-entropty loss in classification and a comparison of kernel machines and deep networks.
Her recent research is about the convergence and generalization of different surrogate loss functions in multi-class classification. She is also broadly interested in optimization or generalization under over parameterized settings.