Research

KAN-Based Implicit Neural Representations

Studied Kolmogorov-Arnold Networks as an alternative architectural prior for implicit neural representations of volumetric scientific data at Argonne National Laboratory. Achieved +2–15 dB PSNR over SIREN/MLP baselines and characterized regimes where learned activation functions outperform fixed-activation INRs via systematic ablation. Trained on the ALCF Sophia DGX A100 cluster.

• Poster Kolmogorov-Arnold Neural Networks as Implicit Neural Representations for Volumetric Data
  Charles O'Hanlon, Mengjiao Han, Joseph Insley, Janet Knowles, Victor Mateevitsi, Michael E. Papka, Silvio Rizzi, Qi Wu (NVIDIA). Learning on the Lawn, Argonne National Laboratory, 2025

Supported by the U.S. DOE Office of Science, Advanced Scientific Computing Research program

[Code] [Poster]

Automated Sit-to-Stand Assessment from Clinical Video

Worked on robustly automating the 30-second chair stand test, a standard sit-to-stand assessment used in physical therapy for older adults, from video. Delivered repetition counts with Mean Absolute Error < 1 repetition vs. clinician annotations across ~50 sessions, comparing end-to-end vs. keypoint-based action-segmentation architectures. Implemented detection and tracking pipelines with MediaPipe, OpenCV, and OpenMMLab models.

Organized by SentiMetrix's PathML; funded by the National Cancer Institute

[Link]

Multimodal Missing Persons Location Prediction

Engineered a large-scale 1M+ multimodal dataset from missing person cases. Prototyped location predictors comparing transformer-based fusion of vision and tabular features against GBDT and agentic baselines.

• Poster Neural Networks for Missing Persons Location Prediction
  Charles O'Hanlon, Brandon Kim, Ameer Arsala, Franz Kurfess. National Missing and Unidentified Persons Conference, 2024

Sponsored by the Menlo Park Fire Protection District

[Code]