Previous | 
Home
Section: Partnerships and Cooperations
International Initiatives
Inria Associate Teams Not Involved in an Inria International Labs
MOHA
-
Title: Mixed Multi-objective Optimization using Hybrid Algorithms: Application to smart grids
-
International Partner (Institution - Laboratory - Researcher):
-
See also: https://ocm.univ-lille1.fr/~talbi/momh/
-
The key challenge of this project is to propose new optimization models and new hybrid algorithms to the demand side management of smart grids in a context of uncertainty and in the presence of several conflicting objectives.
Those complex optimization problems are also characterized by the presence of both continuous and discrete variables. We need to design new efficient optimization algorithms combining state-of-the-art exact and metaheuristic algorithms from the global optimization and combinatorial optimization communities
s3-bbo
-
Title: Threefold Scalability in Any-objective Black-Box Optimization (s3-bbo)
-
International Partner (Institution - Laboratory - Researcher):
-
See also: http://francejapan.gforge.inria.fr/doku.php?id=associateteam
-
The main scientific goals of this collaboration is to theoretically derive, analyze, design, and develop scalable evolutionary and other stochastic local search algorithms for large-scale optimization considering three different axes of scalability: (i) decision space, (ii) objective space, and (iii) availability of distributed and parallel computing resources. This research will allow us to design, control, predict, analyze and optimize parameters of recent complex, large-scale, and computationally expensive systems, providing the basic support for problem solution and decision-making in a variety of real world applications. For single-objective continuous optimization, we want to theoretically derive variants of the state-of-the-art CMA-ES with linear time and space complexity scalings with respect to the number of variables. We will exploit the information geometry framework to derive updates using parametrization of the underlying family of probability distribution involving a linear number of components. The challenges are related to finding good representations that are theoretically tractable and meaningful. For the design of robust algorithms, implementing the derived updates, we plan to follow the same approach as for the design of CMA-ES. For multi- and many-objective optimization, we will start by characterizing and defining new metrics and methodologies to analyze scalability in the objective space and in terms of computational resources. The first challenge is to accurately measure the impact of adding objectives on the search behavior and on the performance of evolutionary multi- and many- objective optimization (EMyO) algorithms. The second challenge is to investigate the new opportunities offered by large-scale computing platforms to design new effective algorithms for EMyO optimization. To this end, we plan to follow a feature-based performance analysis of EMyO algorithms, to design new algorithms using decomposition-based approaches, and to investigate their mapping to a practical parallel and distributed setting.