IDEQ -- Improving Diffusion Models for the Traveling Salesman Problem (TSP) by Leveraging the Structure of the Solution Space
Researchers have developed IDEQ, a novel diffusion model designed to tackle the Traveling Salesman Problem (TSP). By incorporating the structural constraints of TSP solutions and refining curriculum learning, IDEQ achieves state-of-the-art performance on synthetic instances and matches leading heuristics on the TSPlib benchmark. The model demonstrates particular strength on larger instances, achieving near-optimal solutions and exhibiting lower variance and better scalability than previous neural network approaches. AI
IMPACT Sets new SOTA for neural-based TSP solvers, potentially impacting logistics and optimization fields.