Video streams are increasingly becoming an everyday part of life. From amateur GoPro clips to robot vision video streams, there is now a wealth of information documented in video. In these instances, especially regarding robotic vision, it has become necessary to be able to quickly retrieve information from the video streams. In order to more efficiently search through video streams, a technique called coresets is used to compress and segment the video. The video is segmented into scenes, which are in turn compressed into a tree that makes it easy to retrieve specific events from the stream. My work will be focused on exploring multiclass feature learning techniques in order to try to optimize segmentation.

Ever since I started reading science fiction, I have been fascinated with artificial intelligence and machine learning. My second semester at MIT, I took the introduction to machine learning class at MIT, and I thought it was so cool that I wanted to learn more about its applications outside of the classroom. Doing a SuperUROP in CSAIL seems like a great opportunity to explore the research aspect of machine learning.