Matthew Hyatt
I'm a PhD student at Loyola University Chicago, where I work on robot learning. I am affiliated with AISEC and the Software Systems Laboratory. My work is supervised by Mohammed Abuhamad and George Thiruvathukal.
Research Interests: I am interested in self supervised methods for imitation learning, continual improvement, and learning from human behavior. While it is common to use these techniques to improve generalization in unseen environments, I intend to improve the capability of robots to become increasingly capable in familiar environments.
I am interested in collaborating with other motivated students. Please reach out if you would like to work together.
If you are a Loyola student and are interested in contributing to my projects, please fill out this form.
News
Publications
Towards Human-inspired Visual Perception Networks
Luke Baumel, Matt Hyatt, Mikayla Cutler, Joseph Tocco, George K. Thiruvathukal, Nicholas Baker
TLDR: We create a novel dataset of skinned 3d models to study differences in human and machine perception.
Robust Source Attribution of Synthetically Generated Western Blot Images
Matt Hyatt, George K. Thiruvathukal, Daniel Moreira
Loyola eCommons, 2023
TLDR: We use convolution layers in fourier space to improve the classification of western blot images synthesized via GAN/diffusion models.
An Empirical Study of Pre-Trained Model Reuse in the Hugging Face Deep Learning Model Registry
Wenxin Jiang, Nicholas Synovic, Matt Hyatt, Taylor R. Schorlemmer, Rohan Sethi, Yung-Hsiang Lu, George K. Thiruvathukal, James C. Davis
ICSE, 2023
Snapshot Metrics Are Not Enough: Analyzing Software Repositories with Longitudinal Metrics
Nicholas Synovic, Matt Hyatt, Rohan Sethi, Sohini Thota, Shilpika, Allan J. Miller, Wenxin Jiang, Emmanuel S. Amobi, Austin Pinderski, Konstantin Läufer, Nicholas J. Hayward, Neil Klingensmith, James C. Davis, George K. Thiruvathukal
ASE Tools Demo, 2022
TLDR: PRIME is a tool that provides a longitudinal analysis of software repositories to better understand the evolution of codebase over time.
