Efficient Graph Coloring with Neural Networks: A Physics-Inspired Approach for Large Graphs
Автор: fondazione-fair
Загружено: 2025-12-09
Просмотров: 13
Описание:
Andrea Cacioppo (Sapienza Università di Roma, FAIR Spoke 5) presents "Efficient Graph Coloring with Neural Networks: A Physics-Inspired Approach for Large Graphs"
This presentation is part of the Virtual Young Poster Session of the FAIR 2025 General Conference.
For more information: https://fondazione-fair.it/general-co...
Abstract:
Graph-Coloring, is a well-known combinatorial optimization problem, that finds numerous applications in different fields. Exact solutions to the graph coloring problem become computationally infeasible for large graphs and specific connectivities.
We propose a GNN-based algorithm, trained leveraging a physics-informed semi-supervised loss in to solve the graph coloring problem. Although the proposed model is unable to obtain a perfect coloring for every graphs, it can be used as an alternative to traditional optimization algorithms, outperforming them in terms of time efficiency and scalability. In order to demonstrate the performances of the GNN optimizer, a comparison between the model and the simulated annealing algorithm in the case of is presented, applied to Erdos-Renyi Graphs.
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке: