In this project, my team addressed the question: can we use audio data to identify and characterize musical influence relationships? Using data from allmusic.com, we constructed a directed graph encoding musical influence. We then performed link prediction on this graph by training a classifier with input features based on graph structure as well as song audio content data. Our positive results are suggestive of the potential of data-driven analyses of music. [code | report]

The ability to predict taxi traffic could be useful to taxi dispatchers and city planners. My group implemented such a prediction system, applying machine learning to predict the number of taxi pickups given a time interval and location in New York City. This was our final project for an artificial intelligence class (CS221) and was inspired by MIT's 2013-2014 Big Data Challenge, which proposed the same problem for taxicabs in Boston. [code | report]

Continuing our research in computer Arimaa, my group created a bot with knowledge of the "game-phase" of a board position in order to evaluate boards and rank moves more effectively. We conducted this research as a 4-person team engaged in an independent study project under Professor Michael Genesereth. [code | report]

My group created a move ranking model for computer gameplay of the strategy board game Arimaa. We applied supervised learning models to rank moves by 'expertness' and trained on archived move data from games played by experts. This work was the result of a 4-person team project for a machine learning class (CS229). [code | report]