At Swarthmore, I’m working with Dr. Ameet Soni on a machine learning model feature ellicitation. In this project, we’re building a classifier which can request additional features about certain datapoints to improve its accuracy; this project is particularly relevant in medical settings, where one might want to request additional medical tests for certain patients to improve diagnostic accuracy. We will be continuing this project over the academic year, and will focus on making a model which is both accurate and explanatory in terms of which additional features it requests.
I was an REU participant at the Big Data Analytics site at Washington University in Saint Louis. With Dr. Ayan Chakrabarti, I designed an efficient deep-learning model for stereo depth estimation in autonomous vehicles. We worked in low-level TensorFLow, and created a model which performs depth estimation on high-resolution images 1.5x faster than state-of-the-art methods. We are continuing our work throughout the next several months, and will be submitting our results for publishing in early 2019.
I took part in an REU at the University of Colorado at Colorado Springs, where I completed independent computer vision research under the guidance of Dr. Jonathan Ventura. The goal of this project was to use convolutional neural networks along with recent super-resolution architecture to localize fluorescent proteins. You can find the final paper from the Summer here.