The Computing Research Association (CRA) announced its 2020 Outstanding Undergraduate Researcher Awards in December, and two Harvey Mudd College students were commended. Ivy Liu ’20 was selected as a finalist, and Daniel Bashir ’20 earned an honorable mention.
Sponsored this year by Mitsubishi Electric Research Labs, the prestigious program recognizes undergraduates at North American colleges and universities who demonstrate outstanding potential in an area of computing research.
Ivy Liu ’20
A joint mathematical and computational biology major, Liu is interested in developing and applying computational methods to facilitate biomedical research. “Integrating computer science with biology has allowed me to see the beauty of theoretical computer science as well as the applications of tools first-hand,” she says.
Liu’s research experience includes working with biology professor Catherine McFadden to test the feasibility of using a particular gene to differentiate species within the coral genus Sinularia; working with computer science professors Ran Libeskind-Hadas and Yi-Chieh (Jessica) Wu to improve a dynamic programming algorithm for phylogenetic tree reconciliation; and using deep learning to predict DNA sequences related to the remodeling of epigenetic marks driven by a carcinogen with Dr. Cristian Coarfa and Dr. Cheryl Walker at Baylor College of Medicine.
Last summer, Liu worked with Dr. Pavel Sumazin at Baylor College of Medicine to develop computational models to infer cell-type-specific expression from bulk tumor expression profiles. This year, Liu is conducting senior thesis research with biology professor Eliot Bush, developing methods to study the evolutionary history of microbes.
“Through these experiences, I have found a love for computational biology, and I look forward to continuing research on fundamental problems as well as developing tools that will aid biomedical research in the long run,” she says.
Daniel Bashir ’20
“The main purpose of my team’s research is to develop a quantitative framework for overfitting and underfitting in machine learning,” says Bashir. “Both of these pitfalls are major issues for anyone interested in using machine learning for practical purposes.”
Bashir says his research, conducted with other members of computer science professor George Montañez’s AMISTAD Lab at HMC, seeks to answer the question, “given a particular learning algorithm and a particular dataset, by how much will my algorithm overfit or underfit the data?” Having a specific, quantifiable answer to this question for any learning algorithm and set of data would allow a researcher to understand whether or not a particular algorithm is appropriate for a specific task.
“I became interested in this research while I was taking a class from Prof. George in machine learning, information theory and search,” says Bashir, a joint computer science and mathematics major. “I got a chance to think about machine learning from an information-theoretic perspective and see the useful and fascinating parallels between the two fields. This perspective is not only interesting theoretically, but also has practical use. There’s a fair amount of general advice on how to identify whether learning algorithms are overfitting or underfitting and how to fix those problems, but I think that a more quantified framework for answering these questions has the potential to help practitioners iterate on solutions to different problems using machine learning in a more principled way.”