• Aucun résultat trouvé

anytime tree searches for operations research

N/A
N/A
Protected

Academic year: 2021

Partager "anytime tree searches for operations research"

Copied!
2
0
0

Texte intégral

(1)

HAL Id: hal-02155357

https://hal.archives-ouvertes.fr/hal-02155357

Submitted on 13 Jun 2019

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

anytime tree searches for operations research

Luc Libralesso, Vincent Jost, Louis Esperet, Thibault Honegger

To cite this version:

Luc Libralesso, Vincent Jost, Louis Esperet, Thibault Honegger. anytime tree searches for operations research. Journées G-SCOP 2019, May 2019, Grenoble, France. �hal-02155357�

(2)

2019

LATEX TikZposter

ANYTIME TREE SEARCH ALGORITHMS FOR

O

PERATIONS

R

ESEARCH

Luc Libralesso

, Encadrants: Vincent Jost, Louis Esperet, Thibault Honegger

Univ. Grenoble Alpes, CNRS, Grenoble INP, G-SCOP, 38000 Grenoble, France

ANYTIME TREE SEARCH ALGORITHMS FOR

O

PERATIONS

R

ESEARCH

Luc Libralesso

, Encadrants: Vincent Jost, Louis Esperet, Thibault Honegger

Univ. Grenoble Alpes, CNRS, Grenoble INP, G-SCOP, 38000 Grenoble, France

How to solve hard problems ?

local improvement

Constructive

Add tree searches from AI

B&B CP Greedy Ant colony BS SA TS LNS VNS TS PR GA ILS PSO

We believe that tree search from AI is suited for operations research

Tree search principles

.

.

.

initial state

one decision

new state

Each problem can be represented by a search tree. We design generic algorithms that explore this tree.

Applications

Saint-Gobain’s Glass cutting challenge

Triangle Width

Brains on chips

1 4 3 2 5 6 d e f a c b

pack glass items (in blue), minimize the waste (gray) from embedded vision systems partnership with Netri

avoid defects (red dots) and many other constraints maximize "white triangle" aims to design devices that mimic brains

We obtained the best final results tree search algorithms GOAL: better studies on Alzheimer’s disease

during the competition obtain the best results we work on using tree search to generate chips

Research Themes

Integration of AI techniques

Combination with other OR techniques

Many algorithms exist in AI integrate dynamic programming (history cuts)

They are not well known in OR integrate pheromones mechanisms from Ant Colony Optimization

We develop a framework in C++ to use them in OR combine with local search techniques (local search, path relinking etc.)

Offline & Online Learning

Use it on other industrial problems

use machine learning to obtain better guides anytime tree search algorithms seem to be suited to solve complex problems

use machine learning to perform heuristic cuts we obtained excellent results on such problems compared to other approaches

Références

Documents relatifs

When designing ML-based phenological studies, research- ers should consider how to integrate output data and metadata with existing databases or data sets, ideally using terms

Les malades sont pris en charge par plusieurs centres : le service de pneumologie et des maladies infectieuses du centre hospitalier et universitaire Tokoin de Lomé et

Willetts, P.. convergence of the research process. Departing from that, we will clarify in more detail how the Analytical Grid can essentially be deployed as a

Regarding the results of the qualitative variables for the main aim (feature removal or conservation) or for intermediate aims (family parts or features) selected predictive models

We now explain how to approximate hard path problems by a usual shortest path problem using our methodology, and take the pricing subproblems of the stochastic VSP and of the

topology of the configurations in the training set and the testing set are identical. Changes in the topology was outside the scope of the study, as, in the design of critical

Nous avons pris comme corpus dans ce travail ; le texte littéraire donc l’apprenant sera amené à rédiger un récit, les apprenants en question sont des lycéens de

Like the face recognition and the natural language processing examples, most works discussing multi-task learning (Caruana 1997) construct ad-hoc combinations justified by a