MIT EECS — Levine Undergraduate Research and Innovation Scholar
Deep Learning for Finance
The finance market is influenced by complex factors, while large volumes of data are generated within short time. The objective of this project is to develop deep-learning based approaches to analyze finance data to uncover predictive features, and help both semi-automatic and fully automatic trading decision making. We will design a part auto-tuning architecture based on recurrent neural network, and apply multiple optimization techniques to provide high predicting accuracy as well as potentially strong predictive ability on live data streams. This system will be implemented in and deployed to a parallel computing environment, and potentially turned to an open source project.
This project attracts me as it involves both math modeling and system considerations. As a Putnam fellow, I have a good foundation in both applied and pure mathematics. I also interned with Amazon Web Service last Summer, where I designed and developed a cloud based system with an online algorithm which analyzes large amount of data. In this research I hope to dive deep into and gain concrete skills in deep learning.