Tag
Proposes GES, a learning-based graph sparsification method for Euclidean TSP that adaptively prunes up to 95% of edges while maintaining solution quality within 1% of optimal, demonstrating strong generalization.