EVENTO
Blending Physics-Based Models with Machine Learning in the Simulation of Complex Systems. Our Own Experience.
Tipo de evento: Seminário LNCC
Resumo: "The use of computer simulation has leveraged engineers and applied scientists' capacity to handle the design and analysis of complex systems. Moreover, such models are essential in many disciplines due to their intrinsic ability to make forecasts, thus enabling analyzing situations where experiments and observations are not available.Unfortunately, in many situations, the need for accuracy requires the use of complex high-fidelity multiphysics models that directly impact the involved computational costs. Different approaches, like reduced-order models, have been developed with partial success to cope with such an endeavor. More recently, the emerging of powerful methods in the realm of Machine Learning enables the design of a widespread set of tools leveraging more convenient (from the cost point of view) computer models to simulate complex systems. Amongst them, surrogates (like Gaussian Processes, Deep Neural Networks, or Support Vector Machines) and Multifidelity models resulting from the fusion of high and low fidelity data produced by models, observation, and controlled experiments.In this talk, three different situations involving our team's research will be presented: Uncertainty Quantification in Seismic Imaging, the combustion of biogas using reduced kinetic models, and the design of hydraulic fracturing."Relembramos da importância do cadastramento da inscrição e a obrigatoriedade de participação por parte dos discentes da Pós-graduação do LNCC pela plataforma Zoom.Para se cadastrar e assistir pelo Zoom, acesse:https://us02web.zoom.us/webinar/register/WN_LUC7HrNHR8GA_evsZ0XknAAviso Legal: As opiniões expressadas neste vídeo são de responsabilidade exclusiva do(s) palestrante(s).
Data Início: 09/11/2020 Hora: 14:00 Data Fim: Hora: 16:00
Local: LNCC - Laboratório Nacional de Computação Ciêntifica - Webinar
Palestrante: Fernando A. Rochinha - PEM/COPPE - PEM/COPPE
