Matrix Factorization and Reparameterization of Neural Networks
I worked as an intern at the Google-Rice Research Experience for Undergrads studying the effects of neural network reparameterization on accuracy and model size. Below is a presentation I gave on my project which received First Place from Google and Rice University judges. I plan to continue working on this project in the fall.
A Survey of Spike Detection Pipelines
During my senior year of High School, I got the chance to work with a statistics professor at NC State University. With another student, I analyzed a dataset on assaults committed in Raleigh, North Carolina, and we submitted our findings on a poster to the UNC-Charlotte Junior Science and Humanities Symposium receiving an Honorable Mention. The following summer, we worked on a project to develop a virtual lab for high school students to illustrate the challenge of variable selection. During my freshman year, I explored and compared pipelines for detecting spikes in time series data.
As sensors become cheaper to build, it is often more prudent to increase accuracy by increasing the number of sensors rather than designing more robust sensors. This pushes the challenge of ensuring accuracy into the data processing component of the system. This project aims to outline a workable pipeline for automatically detecting upward spikes in time series data.
As sensors become cheaper to build, it is often more prudent to increase accuracy by increasing the number of sensors rather than designing more robust sensors. This pushes the challenge of ensuring accuracy into the data processing component of the system. This project aims to outline a workable pipeline for automatically detecting upward spikes in time series data.