I am a third year PhD student at the Johns Hopkins University in the department of Applied Mathematics and Statistics. I am advised by Soledad Villar. Before coming to Johns Hopkins, I was a Senior Software Engineer at the foodservice startup Cut+Dry. Prior to that, I was trawling the mean streets of Troy, NY, getting a B.S. in Computer Science and Mathematics at Rensselaer Polytechnic Institute, and generally being a nuisance.
My research interests are in the mathematical underpinnings of data science and machine learning. Recently I have been interested in using equivariance to build better models, and dodging the curse of dimensionality through linear and nonlinear dimensionality reduction techniques.
Outside of work I enjoy playing volleyball, watching movies, and reading the news. You can find me at wgregor4 (at) jhu.edu, my office S430 in the Wyman Park Building, or simply prowling the halls of Wyman looking for snacks.
Scalar image convolved with scalar and vector filters
GeometricImageNet: Extending convolutional neural networks to vector and tensor images Wilson Gregory, David W. Hogg, Ben Blum-Smith, Maria Teresa Arias, Kaze W. K. Wong, Soledad Villar
arxiv.org/abs/2305.12585 Preprint, 2023
A model that extends convolutional neural networks to work with images of vectors and tensors. With a simple adjustment, the GI-Net can be made equivariant to changes of coordinates.
MarkerMap: nonlinear marker selection for single-cell studies Nabeel Sarwar, Wilson Gregory, George Kevrekidis, Soledad Villar, Bianca Dumitrascu
arxiv.org/abs/2207.14106 (In Review), 2022
A generative model for selecting minimal gene sets which are maximally informative of cell type origin and enable whole transcriptome reconstruction.