Convite para Seminário: “Modelling and global optimization with splines”
Estamos recebendo a visita de Bjarne Grimstad, pesquisador da NTNU (Universidade de Ciência e Tecnologia da Noruega) que estará defendendo o doutorado em breve. Ele vai estar em visita ao DAS na semana de 21 a 25/Setembro.
Deixo o convite para um seminário em 21/09 (segunda-feira) das 14 h às 15 h na sala PPGEAS-2. Detalhes abaixo,
Prof. Eduardo Camponogara
–
Title of talk: Modelling and global optimization with splines
Abstract: This talk addresses the use of splines to model and optimize black-box functions. Splines, being piecewise-defined by polynomials, are highly flexible functions that may accurately approximate most shapes. A spline can be used to approximate any sampled function, also when the sample is noisy, in which case a smoothing spline should be applied. Splines have many properties attractive for optimization: a high degree of smoothness, analytical derivatives, and a readily available convex hull. These properties are utilized in the recently developed solver CENSO for global optimization with spline constraints. CENSO contains a spatial Branch-and-Bound algorithm that accepts spline constraints modeled with the function approximation software SPLINTER. In this talk, we will introduce the open-source software SPLINTER and CENSO. Examples will be given that demonstrate the flexibility of modeling optimization problems with spline constraints.
Autobiography: Bjarne Grimstad is a PhD candidate at the Department for Engineering Cybernetics, Norwegian University of Science and Technology. He has worked on the topics of function approximation, integer programming, black-box optimization, and global optimization. The motivation behind his work has been to optimize the production from multiphase flow networks by applying various combinations of tools from the above topics. Mr. Grimstad was awarded a PhD scholarship from IBM in 2012 and has since developed the open-source software SPLINTER and CENSO.
