Kinetics-Informed Neural Networks

Authors: Gabriel S. Gusmao, Adhika P. Retnanto, Shachi C. da Silva, Andrew J. Medford

Year: 2020 | Journal: Catalysis Today | Citations: 36+

KINNs embed ODE-based kinetic constraints directly into neural network training. Instead of learning dynamics from data alone, the governing kinetic equations act as physics-based loss terms — enabling accurate estimation of rate constants and activation energies from sparse, noisy transient data where conventional regression fails.

Try It

Kinetics-Informed Neural Networks Playground — fit kinetic models to data in your browser.

Links

arXiv · Catalysis Today · Semantic Scholar · Code

← Back to all publications