Bryan He, a senior student advised by Yisong Yue, and William Hoza, a junior student advised by Leonard Schulman, are the recipients of the 2015 Bhansali Prize. Nicholas Schiefer, a junior student advised by Erik Winfree, won an honorable mention for the Prize. The Bhansali Prize is typically awarded to one undergraduate student for outstanding research in Computer Science in the current academic year, but due to the number of particularly high-caliber candidates in 2015, the Bhansali Prize committee determined that multiple students deserved the award.
Bryan He's research is in the area of interactive machine learning with submodular functions. Bryan has been extending the interactive submodular set cover (ISSC) framework to be more robust to mis-specifications of the model class. His work is largely theoretical, and involves defining new extensions of the ISSC framework, extending existing algorithms in interesting ways, proving optimality guarantees, and running simulation experiments. Bryan's most recent direction is extending his existing results to be able to work with real-valued objective functions. This is an exciting direction, since submodular set cover is difficult to analyze in the non-rational regime. Bryan has shown an unbounded ability to find connections across a variety of theoretical tools to tackle this problem.
William Hoza worked with Professor Schulman on the problem of how to implement distributed protocols, despite adversarial channel errors, on synchronous-messaging networks with arbitrary topology. This problem lies within the broader "interactive coding" field that has drawn attention in CS and EE during the last two decades, and especially the last five years. William identified the problem himself during Winter '14 (as a sophomore), and then worked on it during a SURF in the following summer. He mastered the existing literature, and then improved on it substantially. The work has already been presented in one major workshop, and is under submission for publication.
In addition to being a TA for several Computer Science classes, Nicholas Schiefer has pursued research in quantum computation and molecular programming. In his work with Professor Erik Winfree, he formulated a new theoretical model for programmable molecular systems that combines aspects of formal chemical reaction networks with abstract self-assembly processes, two very well-studied models. Surprisingly, this new model does more than obtain the best of both worlds: it is better at computation than standard chemical reaction networks, and it is better at self-assembly than standard self-assembly models. Nicholas formalized these intuitions with rigorous proofs, including tight bounds on the program-size complexity of self-assembled shapes. A paper, first-authored by and written by Nicholas, has been submitted to the premier conference of the field. According to Professor Winfree, Nicholas is "among the most remarkable undergraduates that I have had the pleasure of working with during my 20-plus years at Caltech.”