MIT EECS Undergraduate Research and Innovation Scholar
Compressing Convolutional Neural Networks for Predicting DNA Methylation
Electrical Engineering and Computer Science
- Machine Learning
David K. Gifford
Sparsifying neural network connections for predicting DNA methylation
This SuperUROP project will develop new visualization formalisms and techniques for both deep learning and hybrid deep/conventional machine learning models. The Gifford laboratory is developing new machine learning techniques to identify DNA sequences that when modified will lead to specific phenotypic outcomes. They have used both conventional Convolutional Neural Networks (CNNs) and approaches that combine CNNs with other classification methods. This project will develop tools to extract meaningful biological signals from these models including motifs and combinatorial grammars that are associated with the observed phenotypic outcomes.
I took part in a research project this summer in machine learning and became really interested in machine learning research and its applications to other fields including but not limited to visualization. This project excites me most because it will give me the chance to apply all the skills I have learned this summer to a project in a field that I find extremely interesting but also challenging