Current member
Landon Harris
PhD Student, Electrical Engineering & Computer Science
Landon works on reinforcement learning and the dynamical-systems view of deep learning. He has been part of AICIP since 2018, starting as an undergraduate researcher on radar–camera fusion for autonomous vehicles.
Landon Harris joined AICIP in fall 2018 as an undergraduate researcher and never really left. His early work sat in the lab’s autonomous-vehicle line — contributing to radar–camera fusion research alongside the CenterFusion effort — before his interests shifted toward the mathematics of learning itself.
He completed his master’s at UT in 2024 with a thesis on the dynamics of diffusion, and continues in the PhD program working on the lab’s Koopman-theoretic research line: understanding neural network training and reinforcement learning as dynamical systems that can be analyzed, predicted, and stabilized. He is a co-author of KIPPO (IJCAI 2025), which brings Koopman-inspired structure to proximal policy optimization.
Projects
Koopman Dynamics
Treating neural network training as a dynamical system — predicting weights epochs ahead to accelerate optimization, and stabilizing policy learning in RL.
EcoCAR
Radar–camera fusion research for connected and autonomous vehicles, and faculty leadership of UT's team in the DOE EcoCAR competitions.
Publications
KIPPO: Koopman-Inspired Proximal Policy Optimization
International Joint Conference on Artificial Intelligence (IJCAI) 2025
Defect Detection in Tire X-Ray Images: Conventional Methods Meet Deep Structures
arXiv 2024