Applied Mathematics

Kyle Loh

I am a fourth-year joint B.S.–M.S. student at Florida Atlantic University studying applied mathematics, where I work with Jason Mireles-James and Yu Xiang.

My research focuses on the mathematical foundations of machine learning, and on using those foundations to build better algorithms.

Google Scholar GitHub LinkedIn YouTube

kloh2023@fau.edu

Research

Rigorous numerics for optimizers

I use computer-assisted proofs to study quasiperiodic and chaotic structures, and routes to chaos, in adaptive optimizers such as RMSProp and Adam. My goal is to understand which mechanisms in an adaptive optimizer drive chaotic behavior.

RMSProp visualizer Adam visualizer

A torus-doubling cascade in RMSProp, shown in four panels: smooth invariant curves break up into progressively more scattered point clouds as a parameter is varied. — A torus-doubling cascade in RMSProp.

Conformal inference

I am developing conformal inference frameworks that bound the false discovery rate under practical constraints: decentralization, limited communication bandwidth, and non-exchangeability.

Publications

Google Scholar

Encoder-decoder schematic mapping an input y through a latent code z to an output x.

Optimal Linear Baseline Models for Scientific Machine Learning

Alexander DeLise, Kyle Loh, Krish Patel, Meredith Teague, Andrea Arnold, Matthias Chung

Foundations of Data Science

Paper arXiv Blog post Code

Four agents, each holding a different local null distribution, shown as four separate Gaussian
point clouds with distinct spreads and orientations.

Decentralized Conformal Novelty Detection via Quantized Model Exchange

Kyle Loh, Yu Xiang

arXiv preprint

Paper

Awards

2026Barry M. Goldwater Scholarship
2026Phi Kappa Phi Scholarship
2026OURI Research Grant
2024Mu Alpha Theta Summer Grant