Personnel: Brady Gess (intern, main developer), Brandon Paul (mentor), Joshua Morgan (mentor), Miguel Zamarripa (supervisor)

Description: Gradient-enhanced neural network (GENN) models use derivatives to inform graph training and are more accurate for multi-input, multi-output problems. Training GENN models requires knowing the derivatives a priori; for a sampled or simulation-produced dataset, these derivatives must be calculated. This project aims to develop a CCSI2 tool to calculates gradients for an arbitrary input dataset to support loading and training GENN models in FOQUS.

Current status: Tool and methods are in-development, and various implementations are being tested for accuracy against a publicly-available sample problem.

Original due date: 6/30
Modified due date: 7/31 or later (end of the summer, probably mid to late August to give Brady time to complete the work)