ProfOptimization2016

Seminar 2026.1

Organizer: Leandro F. Prudente. Talks are held in the Lecture Room of IME/UFG unless otherwise stated.

March 3, 2026

Formulações matemáticas para o problema de escalonamento de tarefas em máquinas paralelas com tempos de setup, famílias de tarefas e com o objetivo de minimizar a soma ponderada dos tempos de conclusão das tarefas

Arthur Kramer · EMSE, France

No problema considerado, um conjunto de máquinas paralelas está disponível para processar um conjunto de tarefas. Cada máquina pode processar apenas uma tarefa por vez e, quando duas tarefas 𝑗 e 𝑘, pertencentes a famílias diferentes, são sequenciadas consecutivamente na mesma máquina, é necessário um tempo de setup entre o processamento das duas tarefas. Este problema é conhecido por ser fortemente NP-difícil e, na literatura, é comumente resolvido por métodos heurísticos e branch-and-bound. Obter soluções ótimas, mesmo para instâncias pequenas, é uma tarefa difícil. Nessa apresentação, cinco novas formulações matemáticas de programação inteira mista serão apresentadas e seus desempenhos analisados por meio de experimentos em mais de 13.000 instâncias disponíveis na literatura.

March 12, 2026

An active-set method for box-constrained multiobjective optimization

Leandro da Fonseca Prudente

We propose an implementable active-set algorithm for smooth box-constrained multiobjective optimization, with emphasis on algorithmic design and computational performance. The method works on one face of the feasible set at a time, treating it as a lower-dimensional region on which the problem simplifies. At each iteration, the algorithm decides whether to remain on the current face or to move to a different one, characterizing two types of iterations: face-exploring and face-abandoning steps. Backtracking and extrapolation strategies are combined, allowing the working set to be expanded or reduced by multiple constraints in a single iteration. Global convergence to Pareto critical points is established, and finite identification of the active set is proved under a dual nondegeneracy assumption, providing theoretical support for the proposed approach. Implementation aspects are discussed in detail, and numerical experiments on standard benchmark problems illustrate the practical effectiveness of the method in comparison with representative deterministic and evolutionary algorithms.

March 19, 2026

Otimizando as configurações de servidores multifuncionais com uma abordagem aprimorada baseada em cenários

Thiago Alves de Queiroz

Este estudo aborda a gestão dinâmica de instalações ambulatoriais, nas quais os servidores atendem pacientes com prioridades distintas e tempos de abandono estocásticos, com diferentes configurações de habilidades. O objetivo é minimizar as penalidades por atraso e abandono, atribuindo dinamicamente configurações aos servidores. A abordagem proposta integra uma busca em vizinhança variável reduzida e recombina soluções de múltiplos cenários para obter soluções eficazes para este problema dinâmico-estocástico. Os resultados em instâncias realistas demonstram que nossa abordagem é mais eficaz do que a abordagem tradicional baseada em cenários e outras políticas de agendamento.

March 26, 2026

A Scalarized Proximal Point Method for Nonconvex Multiobjective Programs

Jurandir de Oliveira Lopes

We propose a scalarized proximal point algorithm for nonconvex multiobjective optimization problems. Using Mordukhovich subdifferentials, we introduce a notion of Pareto–Mordukhovich criticality and show that any accumulation point of the generated sequence is Pareto–Mordukhovich critical under mild assumptions.

April 9, 2026

A Gradient Projection Method with Backtracking on Hadamard Manifolds

Alejandra Muñoz González

In this talk, we consider a constrained optimization problem on a Hadamard manifold, where the objective function is differentiable and the feasible set is closed and geodesically convex. We present a Riemannian gradient projection method with backtracking step size, based on an intrinsic gradient step, a metric projection onto the feasible set, and an Armijo-type line search along a feasible geodesic. We discuss the well-definedness of the method and its asymptotic convergence properties, showing that every accumulation point is stationary under suitable assumptions

April 16, 2026

An adaptive line-search-free multiobjective gradient method and its iteration-complexity analysis

Max Leandro Nobre Gonçalves

In this talk, we introduce an Adaptive Line-Search-Free Multiobjective Gradient (AMG) method for solving smooth multiobjective optimization problems. The proposed approach automatically adjusts stepsizes based on steepest descent directions, promoting robustness with respect to stepsize choice while maintaining low computational cost. The method is specifically tailored to the multiobjective setting and does not rely on function evaluations, making it well suited for this scenario. The proposed algorithm admits two variants: (i) a conservative variant, in which the stepsize is monotonically decreasing; and (ii) a flexible variant, which allows occasional increases in the stepsize. From a theoretical standpoint, under standard Lipschitz continuity assumptions on the gradients, we establish iteration-complexity bounds for achieving a Pareto critical point for both variants in the nonconvex setting In the convex setting, we further derive improved iteration-complexity bounds for the conservative AMG variant. From a practical standpoint, the numerical experiments demonstrate that the flexible AMG performs favorably compared to the steepest descent method with either a fixed stepsize or Armijo line search.

April 23, 2026

Proximal Methods on Hadamard Manifolds

Orizon Pereira Ferreira

This talk will discuss recent ideas related to proximal methods in the setting of Hadamard manifolds. The presentation will address geometric and variational aspects that motivate this method in optimization, highlighting some theoretical properties and possible research directions.

April 30, 2026

Subgradient Methods on Hadamard Manifolds with Lower Curvature Bounds

Jurandir de Oliveira Lopes

The subgradient method is a classical and foundational approach in non-smooth convex optimization, holds enduring importance due its simplicity, robustness, and pivotal role as a conceptual and algorithmic point of departure have made it the backbone of many significant optimization algorithms. Motivated by classical Euclidean results and recent advances in first-order Riemannian optimization, we study the convergence of the subgradient method on Hadamard manifolds with lower curvature bounds. Assuming a nonempty solution set and employing a corresponding non-summable diminishing step-size condition, we establish the convergence of the generated sequence $\{x^k\}$ to a minimizer whenever at least one of the following holds: (a) the sequence $\{x^k\}$ is bounded; (b) the solution set $S$ is bounded; or (c) the step-sizes are square-summable ($\sum_{k=1}^{\infty}\lambda_k^2<\infty$). Additionally, we prove that if $\operatorname{int}(S)\neq\emptyset$, the method achieves finite termination. Our main contribution provides a Riemannian counterpart to Shepilov's Euclidean analysis [Cybernetics, 12 (1976), pp. 544–548], thereby complementing existing literature on convex minimization over manifolds with lower curvature bounds.

May 7, 2026

Modelo de programação por restrições para um problema real de alocação de disciplinas

Layane Rodrigues de Souza Queiroz

Este trabalho aborda o problema de alocação de disciplinas em salas de aula de uma universidade federal localizada no interior de Goiás. O problema consiste em minimizar a soma da carga horária não alocada e o número de salas destinadas a cada disciplina, garantindo uma alocação eficiente e consistente com restrições institucionais, como de demanda, disponibilidade e acessibilidade. O objetivo é comparar dois modelos para a solução do problema, sendo um deles de programação linear inteira, proposto recentemente na literatura, e o outro é de programação por restrições, proposto no presente artigo. Ambos os modelos são implementados computacionalmente e resolvidos com o IBM ILOG CPLEX Optimization Studio. Os experimentos numéricos foram realizados com 36 instâncias, considerando diferentes configurações de demanda de disciplinas e disponibilidade por salas. Os resultados indicam que o modelo da literatura apresenta melhor desempenho na maioria das instâncias, alcançando soluções ótimas em mais instâncias. Apesar disso, o modelo proposto se destaca por apresentar soluções com menos disciplinas não alocadas. Desta forma, a contribuição deste artigo recai sobre a escolha adequada de métodos de solução, neste caso, modelos matemáticos, que possam suportar os tomados de decisão na gestão de recursos em instituições de ensino superior.

May 28, 2026

Convergence Properties of Proximal-Gradient-Type Methods on Hadamard Manifolds

Claudemir Rodrigues Santiago

In this seminar, we will address composite optimization problems on Hadamard manifolds, in which the objective function $F=f+g$ is the sum of a smooth term, not necessarily convex, and a convex term, not necessarily differentiable. We will present an intrinsic proximal-gradient method that solves a subproblem directly on the manifold, avoiding the use of the tangent space or a Euclidean embedding, and we will investigate convergence properties without assuming any Lipschitz continuity condition on the gradient of the differentiable component. We will present three different types of line-search strategies: a monotone descent strategy, a nonmonotone strategy with a max-type scheme, which monitors the maximum value of $F$ attained in recent iterations, and an explicit monotone-type strategy, with greater emphasis on the latter during the seminar.

June 2, 2026

$\epsilon$ projection-based methods for monotone variational inequality problems

Di Liu · FGV

In this talk, my coauthors and I incorporate the $\epsilon$-projection, viewed as a practical realization of the $\epsilon$-enlargement of the normal cone, into three projection-type algorithms for solving monotone variational inequality problems. The proposed schemes replace exact projections with feasible inexact projections governed by relative error criteria. We establish their global convergence and convergence-rate results under suitable assumptions. Numerical experiments further demonstrate that the proposed algorithms with $\epsilon$-projections outperform their exact-projection counterparts in terms of computational efficiency.

June 11, 2026

Asymptotic Convergence of a Projected Gradient Method on Riemannian Manifolds

Maurício Silva Louzeiro

In this seminar, the asymptotic convergence analysis of a projected gradient method on Riemannian manifolds whose sectional curvature is bounded below by a constant will be presented. The method is applied to the problem of minimizing a geodesically convex function over a closed and geodesically convex subset, possibly unbounded. The full convergence of the sequence generated by the method is established when an exogenous step-size rule is employed. In addition, the challenges related to the analysis of the method under other step-size selection strategies are discussed.

June 18, 2026

Title: Inexact Cubic Regularization Method with Adaptive Reuse of Hessian Approximations

Vilmar Gehlen Filho

This work introduces an inexact cubic regularization method with adaptive reuse of Hessian approximations to solve general nonconvex optimization problems. In the proposed approach, the gradient is computed inexactly and updated at every iteration, whereas the Hessian approximation is updated at a specific iteration and then reused for $m$ subsequent iterations (a lazy strategy), where the value of $m$ may vary throughout the procedure. We will present some numerical experiments conducted on the Moré-Garbow-Hillstrom test collection and compare the performance of the proposed method with that of an existing lazy cubic regularization algorithm. The results illustrate the effectiveness of the adaptive Hessian reuse strategy in reducing function and gradiente evaluations while maintaining robustness and efficiency.