Deterministic global optimization over trained Kolmogorov-Arnold networks (KANs)

Deterministic global optimization over trained Kolmogorov-Arnold networks (KANs)

The repository "optimization-over-KANs" provides resources for deterministic global optimization over trained Kolmogorov-Arnold networks (KANs), which can serve as surrogate models in process optimization.

The repository includes mixed-integer nonlinear programming (MINLP) formulations of trained KANs implemented in Pyomo, Python scripts for training multilayer perceptrons (MLPs) with TensorFlow for comparison, and optimization routines for trained MLPs using OMLT. It also contains the code and data needed to reproduce the results of the associated paper. For the research context, see the project page Optimization with machine learning models embedded.

Open-source code

The source code is available on GitHub.

Key publications

  • Karia, T., Lastrucci, G., & Schweidtmann, A. M. (2025). Kolmogorov Arnold Networks (KANs) as surrogate models for global process optimization. Systems and Control Transactions (Proceedings of ESCAPE 35). doi:10.69997/sct.195815 · Details
← Back to Software