arxivphysics.comp-phcs.LG2026-07-02
CodeJeNN: A simple C++ neural network generator for physics applications
Jay Arcities, Pavel Popov, Eric J Ching, Kamal Viswanath, Ryan F Johnson
Machine learning has shown speedups for numerical methods in physics applications, but integrating Python-based libraries into high-performance C++ solvers creates performance bottlenecks. We present CodeJeNN, which bridges this gap by auto-generating self-contained C++ code from…