Lab Home | Phone | Search
Center for Nonlinear Studies  Center for Nonlinear Studies
 Colloquia Archive 
 Postdoc Seminars Archive 
 Quantum Lunch 
 CMS Colloquia 
 Q-Mat Seminars 
 Q-Mat Seminars Archive 
 Kac Lectures 
 Dist. Quant. Lecture 
 Ulam Scholar 
 Summer Research 
 Student Application 
 Past Visitors 
 PD Travel Request 
 History of CNLS 
 Maps, Directions 
 CNLS Office 
Monday, September 16, 2019
3:30 PM - 4:15 PM
CNLS Conference Room (TA-3, Bldg 1690)


Combining symbolic and numeric AI for modeling: Intelligent formal methods for complex model stacks

Eric Mjolsness
University of California, Irvine

Progress in modeling complex, heterogeneous, spatially distributed, multiscale biological systems is being made by applying both machine learning methods (numerical AI - Artificial Intelligence) and declarative mathematical modeling languages including computer algebra (symbolic AI). I will use as examples reaction/diffusion networks relevant to modeling synapses during learning, and current work on cell-scale microtubule networks in plant development. Technical tools include “dynamic Boltzmann distributions” for model reduction by machine learning, and “dynamical graph grammars” for modeling evolving, spatially embedded structures. I will then attempt to set the stage for further discussions on AI for Computational Science by considering formal representations of computable scientific models and their stackable, conditionally valid reduction mappings, within a conceptual architecture for accumulating and applying computer-represented knowledge of the relevant applied mathematics.**This seminar is part of a series on Artificial Intelligence for Computational Science.

Host: Aric Hagberg