AI-Guided Adaptive Multiscale Modeling of Platelet Dynamics
Event Type
ACM Student Research Competition: Graduate Poster
ACM Student Research Competition: Undergraduate Poster
Posters
Student Program
TP
XO
TimeWednesday, 18 November 20208:30am - 5pm EDT
LocationPoster Module
DescriptionWe developed an AI-guided adaptive multiple time stepping algorithm to model platelet activation, adhesion and aggregation, complex dynamical processes that cause physiological reactions including cardiovascular diseases and stroke. The dynamics spans 6 spatial and 9 temporal scales. Our algorithm can intelligently adapt integration time step sizes to the underlying dynamics. We access the accuracy and speed of our algorithm on a heterogeneous supercomputer with the IBM POWER9 CPUs and Nvidia V100 GPUs. The algorithm speed increases by 4000x with CPUs and an extra 5-10x with GPUs while preserving the solution accuracy within 97%, compared with the conventional algorithm. The poster presents the details of the AI-guided multiple time stepping algorithm and its performance for speeding up modeling of this challenging multiscale biomedical problem.
