Tackling discretised PDEs with a novel AI-based approach, the study Solving the Discretised Neutron Diffusion Equations using Neural Networks relies on the tools within AI software libraries.

• Introduces a new method for representing numerical discretisations through neural networks.
• Demonstrates the equivalence of network solutions to those from traditional codes.
• Benefits from AI libraries include optimized execution on CPUs, GPUs, and new AI processors.
• Uses a U-Net architecture to simulate a sawtooth multigrid solver for reactor physics applications.

This creative use of AI libraries extends beyond the typical applications into the domain of numerical solutions for scientific problems. This research not only proposes an innovative approach to a long-standing problem but also opens doors for other scientific disciplines to adopt such AI-driven methodologies.