The Evolutionary Design and Optimization Group at CSAIL is working on the first large-scale genetic programming (GP) system supported by the cloud. The system, called FlexGP, spans several computational substrates, providing elastic scalability and flexible factorization across multiple dimensions. My project will be to try to achieve higher speeds in evaluation of models used by FlexGP. Specifically, I will study how the application of coresets to our algorithm can improve its speed and performance, and try to answer questions such as: what are the benefits of coresets? How do we exploit coresets maximally depending on FlexGPs goals? How much computation does forming coresets require? How does the process scale?

In my last UROP I analyzed the effects of using bootstrapping techniques to improve the performance of FlexGP. This past summer I worked with the big data processing team at Microsoft, analyzing the differences between their Machine Learning library and Mahout (Hadoops version of machine learning). I am very interested in large scale computing and machine learning.