Robert Nishihara is a 3rd-year PhD student at UC Berkeley working on machine learning and optimization with Michael Jordan. His research focuses on building tools and theory to effectively solve machine learning problems.
As optimization lies at the heart of much of machine learning, some of his work has focused on designing optimization algorithms capable of taking advantage of problem structure and on understanding the behavior of optimization algorithms in a variety of settings (e.g., the combinatorial setting, convex setting, and hyperparameter optimization setting). On the systems side, he has worked on building practical systems for the distributed training of deep networks, and he is currently working on building more general systems for distributed computation.
Some of his more recent interests include creating algorithms capable of learning using less labeled data and for inferring causal relationships between variables.