Workshop: ISAV26: In Situ AI, Analysis and Visualization
Authors: YEHUA YANG (University of Macau); YI JIANG and RONGLIANG CHEN (Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences); and XIAO-CHUAN CAI (University of Macau)
Abstract: Cardiac arrhythmias are associated with abnormal propagation of electrical activity through the heart. High- resolution simulations can capture these dynamics but generate large data volumes that are expensive to store and transfer. Fixed output intervals may save redundant states while undersampling periods of rapid electrical propagation. We propose an adaptive in situ workflow that learns when simulation states should be retained. An offline teacher identifies informative voltage states, and a lightweight causal neural network is embedded directly in the parallel simulation to make runtime output decisions. The selected nonuniform data are then used to train a continuous-time neural operator. Compared with uniform temporal sampling, the adaptive strategy stores fewer states while improving prediction accuracy on unseen simulations. The results demonstrate an in situ connection between high-performance cardiac simulation and neural operator learning.
Back to ISAV26: In Situ AI, Analysis and Visualization Archive Listing
Back to Full Workshop Archive Listing