Neural Control of Attention and Decision Making
At any given moment, activity within our brain is constantly changing. However these changes are not random, but structured with certain spatial and temporal dynamics. For one of my thesis projects, I investigated how these brain dynamics were related to changes in attention: simultaneously recording fluorescent changes in the mouse cortex as an approximate measure of voltage and the mouse pupil which has been shown to fluctuate with changes in attention and arousal. Our results demonstrate that changes in pupil diameter are able to account for up to 70% of the variance in the brain activity, demonstrating just how important arousal and attention changes are in influencing ongoing brain activity. Data and analysis tools will be available through the github and google storage links below.
TransitML - Advanced Analytics in Transportation
Public transportation is a critical piece of infrastructure for cities across the world. Facilitating the use and optimizing the function of these transportation systems can provide tremendous benefit for the citizens of these cities. To aid in this effort, many cities have open sourced historical data or provided API utilities to acquire the data in real time. Through this project we ultimately aim to provide a web or mobile interface to examine historical performance and descriptive statistics in addition to building improved models for service disruptions and arrival predictions. While initially started to examine Atlanta's MARTA transit system specifically, the scope has now expanded to building bus, rail, and bike-sharing datasets in major cities across the US. All code and analysis tools are currently available through github. Published cleaned datasets are provided through google storage links below. *Please note that I do not own the data provided. All usage must comply with the rules associated with the respective city agencies*.
Latent Factors in Financial Markets
The ability to accurately classify the state of financial market activity provides improved insight in to market health and stability as well as improved modelling and forecasting accuracy. The behavioral states of financial markets can be studied similar to one analyzes human or animal behavior and phyiology. While measures like stock price volatility and correlation structures have been previously used to classify market states, recent advances in deep learning techniques can account for non-linear relationships in market activity and extract the latent or "hidden" factors underlying market behavior. Sequential auto-encoders are deep neural networks that can be trained to compress the input data to a smaller set of factors (encoder network) and reconstruct the original data from the compressed factors (decoder network). With this method we are able to create a reduced set of factors that account for the changes in financial market activity and cluster the relationships between these factors to identify market states. Data and analysis code will be released soon at the links below.
