Stability-Informed Neural ODEs
||Stability-Informed Neural ODEs|
||Erik Frisk <firstname.lastname@example.org>|
||2023-10-31 – 2024-05-01|
Neural Ordinary Differential Equations (neural ODEs) offer an attractive approach to modeling dynamic systems. However, optimizing their training remains challenging. This project seeks to utilize the supercomputing resources to intensively investigate the relationship between numerical integration methods, stability regions, step size variation, and model initialization in neural ODEs across various data sets and ML benchmarks.