New course provides research experience for undergraduate students

Understanding how research works and being able to propose and execute research plans are valuable skills for students that enhance their critical thinking and lead to a positive scientific identity. Creating opportunities for undergraduate students to become involved in research and gain this experience is a priority for Dr. Sindhu Kutty, a faculty member at the U-M Computer Science and Engineering Division (CSE) and director of CSE’s Teaching Lab.

To create a new opportunity for students looking for an introduction to research, Kutty introduced a new course in Fall 2024, EECS 498: Machine Learning Research Experience. The course is for undergraduate students who have taken an upper-level machine learning course and are interested in pursuing curious independent research.

Machine learning is a branch of artificial intelligence that is focused on enabling computers and machines to imitate the way that humans learn, to perform tasks autonomously, and to improve their performance through experience and exposure to more data.

“This course is designed to take students step-by-step through the process of doing research,” said Dr. Kutty. “I try to provide both autonomy and scaffolding so that students can have a positive experience while building their understanding of the research process.” Kutty adds that one part of this skill is the technical component and the other part is research communication. Kutty works closely with students to provide feedback as they build their skills in both these dimensions. 

A major component of the course is a semester-long replication project. Students initially conduct a broad literature survey on various applications of machine learning, including natural language processing, reinforcement learning, vision, and fairness. After that, students work in groups of 4-6 students to explore a paper of their choice in detail and submit a project proposal that identifies a significant slice of the paper that they can reasonably replicate in the remaining time and propose an extension to the state of the art that they can implement and test.

“Dr. Kutty’s Machine Learning Research Experience course was valuable in deepening my understanding of the research process,” said undergraduate data science student Minjia Tang. “The course covered all aspects of research, including efficiently reading literature, replicating previous work, giving deliveries, and more. Each group then applied what we learned by working on our own replication projects. Throughout the course, we had regular individual meetings with Dr. Kutty, who provided helpful guidance and advice. With her support, our group successfully completed a project that used deep learning models to restore old photographs.”

At the end of the course, the students presented about their work in a lightning talk and submitted a conference-style term paper. They also presented to the public at a poster session held in Tishman Hall, the atrium of the Bob and Betty Beyster Building.

“Introduction to Machine Learning Research has been one of the best courses I have taken here at U-M,” said undergraduate computer science student William Zheng. “Prior to taking this course, I had limited experience with research in general, much less than machine learning research. After taking this course, I felt I had gained a lot of valuable experience in what it meant to do actual research in the machine learning field.

“From this course,” he continued, “I came to understand not just technical details of the field, such as learning how to run ML inference, but the culture and expectations that come in the field. I would highly recommend anyone looking to get into machine learning research to take this course, especially if lacking experience.”