Paul Jason Mello
PhD Student
Computer Science and Engineering
"All is flux."
- Heraclitus
Research Interests
- Machine, Reinforcement, and Deep Learning.
- Information Theory and Generative Modeling.
- Categorical, Geometric, and Manifold Methods.
About Me
Hi, I'm Paul! A PhD student at the University of Nevada, Reno researching the deep mathematical foundations of deep learning.
My research interests bridge topics like information theory, category theory, and manifold methods to understand and improve how neural networks learn.
Recently I have completed work in building a deep learning framework from categorical first principles and studying delayed generalization through manifold geometry.
For collaborations or questions, feel free to reach out via my email or socials.
Projects
-
Categorical Deep Learning
A deep learning framework built from categorical first principles.
-
Transformer Attention Bench
Benchmarking suite comparing attention variants across efficiency and accuracy.
Publications
-
An Exploration of Information Processing in Diffusion Models
San José State University · M.S. Thesis
Information-theoretic analysis of denoising diffusion models.
-
Multi-Resolution Diffusion for Privacy-Sensitive Recommender Systems
IEEE Xplore
Multi-resolution diffusion for privacy-sensitive recommender systems.
Writings
-
Echo State Networks
Research on echo state networks and reservoir computing.
-
Liquid Time-Constant Networks
Research on continuous-time recurrent networks with adaptive dynamics.
Education
- PhD in Computer Science and Engineering - University of Nevada, Reno Fall 2024 - Present
- MSc in Artificial Intelligence - San José State University Fall 2021 - Spring 2024
-
BSc in Computer Science - California State University, Sacramento
Fall 2016 - Spring 2021
- Minors: Mathematics, Philosophy