Multi-Fidelity Learning with Shallow Recurrent Decoders for Reactor Physics
Researchers have developed a new machine learning architecture called Shallow Recurrent Decoders to improve the accuracy of reactor physics simulations. This method maps low-fidelity data, such as that from point kinetics models, to higher-fidelity solutions like the diffusion equation. By leveraging multi-fidelity information, the technique significantly reduces computational costs while providing more detailed insights into neutron behavior. AI
IMPACT This research could lead to more efficient and accurate simulations in nuclear engineering and other complex scientific fields.