Data e horário: Segunda-feira - 14/09 - 09h30
Local: Auditório - Bloco V
Chair: Soledad Aronna
Título: Adaptive Time Integration Methods, Software, and Applications
Resumo: Adaptive time integration has become an essential tool for the efficient and accurate simulation of time-dependent science and engineering applications. In this presentation, we develop the formulation of both single-rate and newly developed adaptive multirate methods, emphasizing how these multirate approaches exploit disparate temporal scales to achieve greater efficiency than standard single-rate schemes. A further focus is on software infrastructure, where robust implementations of adaptive and multirate integrators enable practical deployment in large-scale scientific computing. We highlight how these methods are hardened in the SUNDIALS software framework and discuss their use in applications spanning quantum dynamics, combustion, and nuclear physics. Together, these examples demonstrate the impact of advanced time integration methods on the simulation of complex physical systems and underscore the interplay between numerical analysis, algorithm design, and scientific software development.
Short-bio: Dr. Carol Woodward is a Distinguished Member of the Technical Staff at Lawrence Livermore National Laboratory and currently serves as President of the Society for Industrial and Applied Mathematics (SIAM). She received her PhD in Computational and Applied Mathematics from Rice University and has been at Lawrence Livermore National Laboratory in the Center for Applied Scientific Computing since then. She is the Director of the Frameworks, Algorithms, and Software Technologies (FASTMath) Institute, a 10-institution effort to provide numerical algorithms expertise and software to scientists across the United States Department of Energy. She also leads the development and deployment of the SUNDIALS package of time integrators and nonlinear solvers which garners over 100,000 downloads/clones each year. Her research interests include numerical methods for nonlinear partial differential equations, nonlinear and linear solvers, time integration methods, numerical software development, and high performance computing. Dr. Woodward was named to the 2017 Class of Fellows of SIAM and to the 2021 Class of Fellows of the Association for Women in Mathematics. In 2023 Dr. Woodward was a co-winner of the ACM/SIAM Prize in Computational Science and Engineering as part of the SUNDIALS Core Development Team.
Data e horário: Terça-feira - 15/09 - 10h30
Local: Auditório - Bloco V
Chair: Douglas S. Gonçalves
Título: Métodos regularizados de descida por coordenadas em blocos: complexidade e aplicações
Resumo: Os métodos de descida por coordenadas em blocos têm sido, há muito tempo, uma ferramenta importante em otimização de grande escala. Mais recentemente, esses métodos receberam atenção renovada devido ao crescente destaque das aplicações em aprendizado de máquina, nas quais a escalabilidade é um requisito fundamental. Nesta palestra, apresentaremos avanços recentes em métodos de coordenadas em blocos com regularização para problemas de otimização com restrições de caixa e problemas de minimização com restrições não convexas. Para problemas com restrições de caixa, serão descritos métodos que empregam regularização de ordem superior e será discutida sua convergência assintótica para pontos estacionários de ordem p. Também serão apresentados resultados de complexidade no pior caso, mostrando que a obtenção de estacionariedade de primeira ordem com precisão ε, em relação às variáveis de cada bloco, requer O(ε−(p+1)/p) iterações, enquanto a obtenção de estacionariedade de primeira ordem com precisão ε, em relação a todas as variáveis, requer O(ε−(p+1)) iterações. Para problemas com restrições não convexas, serão considerados métodos de coordenadas em blocos baseados em regularização quadrática. Serão discutidas suas propriedades de convergência e a análise de complexidade associada, mostrando que um ponto (δ, ε)-crítico pode ser obtido com O(ε−2) iterações e avaliações da função. Será ilustrado o desempenho prático dessas técnicas por meio de aplicações recentes em problemas emergentes de roteamento motivados por entrega de última milha (last-mile delivery) e operações com veículos aéreos não tripulados (unmanned aerial vehicles — UAVs).
Data e horário: Quinta-feira - 17/09 - 10h30
Local: Auditório - Bloco V
Chair: Carlile Lavor
Título: Random projections in mathematical programming: recent advances
Resumo:Random Projections (RP) provide a probabilistic guarantee of approximate congruence between two point sets having different dimensions. For this reason, they are frequently used as a dimensionality reduction method. I will talk about the application of RP to Mathematical Programming (MP), which is the formal language of optimization. Over the past decade, we have successfully applied RP to several MP subclasses: Linear Programming (LP), conic programming including Semidefinite Programming (SDP), and some Quadratic Programming (QP) variants. We also applied RP to individual problems: the Distance Geometry Problem (DGP), the Alternating Current Optimal Power Flow (ACOPF) problem, and the k-means problem also known as the Minimum Sum-of-Squares Clustering (MSSC) problem. I will review this material, then sketch a new application of RP to Quadratically Constrained Quadratic Programming (QCQP), as well as a new type of application of RP to the MSSC.
Data e horário: Sexta-feira - 18/09 - 11h30
Local: Auditório - Bloco V
Chair: Luiz-Rafael Santos
Title: Brazil's Market of Natural Gas: Explicit Reformulations of Complementarity Equilibria
Abstract: In mathematical optimization, finding an equilibrium for competitive markets involves dealing with complementarity models. When agents in the market have an elastic behavior, complementarity problems are very challenging and hard to solve. To address this difficulty, we propose an equivalent reformulation that maps complementary equilibria into solutions of ractable optimization problems.
The computational efficiency and robustness of the proposed reformulation is assessed on the Brazilian natural gas transportation network.
Joint work with Marcelo M. Cordova, Juan Pablo C. Luna, and Laura Bahiense
Data e horário: Terça-feira - 15/9 - 11h30
Local: Auditório - Bloco V
Chair: Sandra Malta
Título: Oncologia Computacional: Abordagens e Desafios
Resumo: A palestra oferecerá uma visão do estado da arte da modelagem do câncer, demonstrando como ferramentas matemáticas e computacionais são usadas para simular a evolução da doença e avaliar estratégias de tratamento. Ao mesmo tempo, fará uma discussão crítica sobre os desafios atuais para tornar esses modelos mais precisos e úteis na prática clínica.
Data e horário: Terça-feira - 15/9 - 11h30
Local: Anfiteatro IV - Bloco V
Chair: Pablo Rodriguez
Títle: From Diffusion Equations to Nonparametric Estimation
Abstract: Diffusion equations play a central role in many areas of applied mathematics as natural smoothing mechanisms. In recent years, this idea has also been incorporated into nonparametric statistics, leading to diffusion-based estimators with strong practical performance. In this talk, we present a theoretical framework for this class of methods in both density estimation and nonparametric regression. We will show how diffusion can be interpreted as a regularization operator acting on an initial estimator and establish uniform convergence results together with explicit sup-norm error bounds. The analysis reveals that the convergence properties of the initial estimator are naturally inherited by its diffused counterpart. We will also discuss two diffusion-based regression estimators, their theoretical properties, and numerical experiments illustrating the practical benefits of the diffusion approach.
Data e horário: Quinta-feira - 17/9 - 11h30
Local: Auditório - Bloco V
Chair: Daniela Buske
Title: When Biology Becomes a Mathematical Problem
Abstract: Biology has undergone a profound transformation over the last two decades. Advances in high-throughput technologies have produced an unprecedented volume of genomic and clinical data. Yet today the primary challenge is no longer data generation. It is understanding complexity. Biological systems are inherently heterogeneous, dynamic, multiscale, and uncertain, making them one of the most demanding problems in modern applied mathematics. Advances in sequencing technologies have transformed biology into one of the most data-intensive scientific disciplines. Public repositories now contain millions of biological datasets, yet the availability of data has not translated proportionally into biological understanding. The central challenge has shifted from data generation to data interpretation. In this talk, I will discuss how modern genomics has become fundamentally a mathematical and computational problem. High-dimensional observations, heterogeneous experimental platforms, temporal dynamics, and intrinsic biological variability demand statistical methodologies, optimization techniques, machine learning algorithms, and probabilistic models capable of extracting robust and reproducible knowledge from complex datasets.
Data e horário: Quinta-feira - 17/9 - 11h30
Local: Anfiteatro IV - Bloco V
Chair: Kelly Poldi
Título: Problemas de Logística & Transportes Inteligentes: Soluções via Otimização & Inteligência Computacional.
Resumo: Nesta apresentação iremos abordar algumas aplicações e soluções de segmentos de Logística e Transportes no contexto de Cidades & Regiões Inteligentes e Sustentáveis (CRIS), bem como apresentar algumas das metodologias mais eficientes existentes na Otimização e Inteligência Computacional para resolver estes desafios, incluindo os chamados: Algoritmos Híbridos. Mostraremos algumas pesquisas desenvolvidas pela nossa equipe do LabIC (Laboratório de Inteligência Computacional) do IC/UFF em temas associadas a CRIS que incluem aplicações em logística humanitária e Eventos Climáticos Extremos (ECE).
Sobre o Palestrante
O Professor Luiz Satoru Ochi é professor titular do Instituto de Computação da Universidade Federal Fluminense (IC-UFF). É bolsista de produtividade em pesquisa do CNPq (Comitê de Computação, nível PQ-1C) e, desde 2008, recebe o Grant da FAPERJ associado ao prêmio Cientista do Nosso Estado (CNE-FAPERJ). Em 2020, recebeu o Prêmio de Excelência Científica pela UFF. É membro do Comitê Gestor do HUB-Rio (Centro de Excelência em Transformação Digital & Inteligência Artificial do Estado do Rio de Janeiro) e atua anualmente na organização e em comitês de diversos eventos nacionais e internacionais nas áreas de Otimização, Computação e Engenharias. Seus temas de ensino e pesquisa incluem: Inteligência Computacional, Inteligência Artificial, Algoritmos, Computação de Alto Desempenho, Pesquisa Operacional e Otimização.
Data e horário: Sexta-feira - 18/9 - 10h30
Local: Auditório - Bloco V
Chair: José A. Cuminato
Title: Recent Advances in Scientific Machine Learning for Coupled Flow and Transport
Abstract: Scientific Machine Learning is an emerging field with significant impact on methods for solving problems in science and engineering, particularly for large-scale coupled fluid-flow and transport problems. Numerical simulations for these problems can be costly, making such approaches valuable for understanding and improving the efficiency of state quantification and prediction. This talk will review recent advances in scientific machine learning, including dynamic mode decomposition, manifold learning, graph neural networks, and neural operators (Fourier Neural Operators, DeepOnet, and hybrid schemes), as applied to coupled fluid-flow and transport problems. These problems are of interest in sustainable resource exploration, geophysics, and various industrial applications. The talk will show how data-driven information can improve the efficiency of numerical simulation software, explore parametric manifolds for unseen scenarios, and reconstruct high-dimensional simulations using lower-dimensional structures in feasible time.
Data e horário: Sexta-feira - 18/9 - 11h30
(CANCELADA)