Journées Julia et Optimisation

Session

JuMP & OR

Oct 5, 2023, 9:00 AM

Jean-Baptiste Say (CNAM Paris)

5. What's new in HiGHS

Julian Hall (University of Edinburgh)

10/5/23, 9:00 AM

HiGHS is the world's best open-source linear optimization software, and has had a good and valuable relationship with the JuMP community for several years. This talk will present the recent developments in HiGHS. In particular it will give a progress report on new interior point solver, and set out the motivation for its development.

14. Building Coluna.jl, a Branch-and-cut-and-price framework in Julia

Guillaume Marques

10/5/23, 9:30 AM

Coluna.jl is a branch-and-cut-and-price framework written in Julia. Together with BlockDecomposition, this package allows you to seamlessly run a branch-and-cut-and-price algorithm from your JuMP model. The platform provides default implementation of advanced features such as non-robust cuts, stabilization techniques that you can easily customize to develop your own next-stage state-of-the-art...

25. Large Scale Optimization via Monte Carlo Tree Search

Larkin Liu (Technical University of Munich)

10/5/23, 10:00 AM

We introduce the application of Monte-Carlo Tree Search (MCTS) for a well-defined large scale Markov decision process. As an example we investigate, the ship refuelling problem subject to subject stochastic fuel prices. As a benchmark model, we compare this MCTS-based implementation to pure stochastic programming approaches with scenario tree generation. The objective of the new approach is to...

15. JuMP into Constraint Programming

Jean-Francois Baffier

10/5/23, 11:00 AM

This endeavor focuses on integrating Constraint Programming (CP) into the versatile JuMP.jl framework. We compile a set of common XCSP3-standard constraints, each tailored to align with a type-based parameter interface, facilitating seamless multiple dispatch. This interface, distinct from JuMP, serves as a unifying structure for these constraints. Exploring classical problems like Sudoku,...

26. From vOptGeneric.jl to MultiObjectiveAlgorithms.jl

Gokhan Kof, Oscar Dowson, Xavier Gandibleux (Université de Nantes)

10/5/23, 11:30 AM

vOptGeneric.jl is a JuMP extension package for modeling and solving
multi-objective linear optimization problems It was created as part of the
ANR/DFG research project vOpt (2015-2019). Inspired by the lessons learned in
vOptGeneric, we added first-class support for multi-objective optimization
problems to JuMP and MathOptInterface. In addition, we developed a...