BEGIN:VCALENDAR VERSION:2.0 PRODID:Linklings LLC BEGIN:VTIMEZONE TZID:America/New_York X-LIC-LOCATION:America/New_York BEGIN:DAYLIGHT TZOFFSETFROM:-0500 TZOFFSETTO:-0400 TZNAME:EDT DTSTART:19700308T020000 RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU END:DAYLIGHT BEGIN:STANDARD TZOFFSETFROM:-0400 TZOFFSETTO:-0500 TZNAME:EST DTSTART:19701101T020000 RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU END:STANDARD END:VTIMEZONE BEGIN:VEVENT DTSTAMP:20210402T160552Z LOCATION:Track 3 DTSTART;TZID=America/New_York:20201113T152500 DTEND;TZID=America/New_York:20201113T154000 UID:submissions.supercomputing.org_SC20_sess222_ws_cafcw131@linklings.com SUMMARY:Machine Learning Driven Importance Sampling Approach for Multiscal e Simulations DESCRIPTION:Workshop\n\nMachine Learning Driven Importance Sampling Approa ch for Multiscale Simulations\n\nBhatia, Consortium\n\nAlmost all phenomen a in science and engineering are inherently multiscale and many require ex ploration across orders of magnitude in both space and time. Solving such problems at the finest scales is computationally prohibitive and, instead, they are often formulated using multiscale models. Coupling the two scale s, however, remains a challenging problem. Here, we present a new automate d way to loosely couple two scales through a Machine Learning (ML) driven adaptive sampling approach that can focus on a user-defined hypothesis, e. g., diversity sampling. Our work advances the paradigm of heterogeneous, m ultiscale simulations by providing a generic framework to couple several s cales in cascade in an arbitrarily scalable manner. We demonstrate our tec hnique on multiscale simulations of the interactions of RAS and RAF protei ns with plasma membrane in the context of cancer-signaling mechanism.\n\nG iven two scales to be coupled, macro and micro, our sampling framework use s a ML-based approach, supervised or unsupervised, to learn the important yet possibly-hidden features by exploring the space of macro configuration s. Using the space of characteristic macro features, the model is able to distinguish between similar and dissimilar configurations. Next, we use a dynamic, adaptive sampling approach in the feature space to identify the m ost important configurations with respect to the scientific hypothesis und er investigation. These selected configurations are promoted to be simulat ed at the micro scale. Given sufficient computational resources, our frame work produces a macroscale simulation that, for each macro configuration e xplored, contains a microscale simulation similar enough to serve as stati stical proxy. Our sampling approach also provides a means to debias the sa mpling through appropriate importance weighting, which allows reconstructi ng the relevant statistics of the microscale using macro samples. As a res ult, our framework is able to deliver macro length- and time-scales, but w ith the insights effectively from the microscale behavior.\n\nOwing to its automated and dynamic approach, our sampling framework is capable of prod ucing massively parallel multiscale simulations that scale to the largest machines on the planet. Previously, we utilized our sampling framework to conduct over 116,000 select microscale simulations, aggregating a total of 200 ms of coarse-grained trajectories, sampled from a 152 s long continuum (macroscale) simulation, utilizing the whole of Sierra with thou sands of CPUs and GPUs for several days.\n\nHere, we present our framework extended to support three scales (continuum, coarse-grained, and atomisti c) of simulations of RAS and RAF proteins on plasma membranes with eight l ipid species. Our framework uses an unsupervised deep learning model, a ta ilored autoencoder, focusing on capturing the spatial response of lipids t o the presence of protein(s) under consideration to select “important” con tinuum configurations. We also present a novel approach to identify “impor tant” coarse-grained configurations in a supervised approach using three b iologically relevant reaction coordinates. Using these importance criteria , we create a three-scale simulation model that investigates the interacti ons between RAS, RAF, and the membrane in progressive detail based on the importance of simulated configurations. We report early breaking results f rom our simulation campaign run on Summit and demonstrate the flexibility, generalizability, and scalability of our framework.\n\nRegistration Categ ory: Workshop Reg Pass END:VEVENT END:VCALENDAR