A Diffusion Model Framework for Unsupervised Neural Combinatorial Optimization
About
Learning to sample from intractable distributions over discrete sets without relying on corresponding training data is a central problem in a wide range of fields, including Combinatorial Optimization. Currently, popular deep learning-based approaches rely primarily on generative models that yield exact sample likelihoods. This work introduces a method that lifts this restriction and opens the possibility to employ highly expressive latent variable models like diffusion models. Our approach is conceptually based on a loss that upper bounds the reverse Kullback-Leibler divergence and evades the requirement of exact sample likelihoods. We experimentally validate our approach in data-free Combinatorial Optimization and demonstrate that our method achieves a new state-of-the-art on a wide range of benchmark problems.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Maximum Independent Set | ER [700-800] | Solution Size43.98 | 58 | |
| Maximum Independent Set | RB [200-300] | MIS Size19.88 | 38 | |
| Maximum Independent Set | MIS Large-scale | Objective Value38.1 | 21 | |
| MaxCut | GSET (|V|=800) | Average Gap to Best Known Cut4.11 | 19 | |
| MaxCut | GSET (|V|=1K) | Average Gap to Best Known Cut6.33 | 19 | |
| MaxCut | GSET |V|=2K | Average Gap to Best Known Cut31.67 | 18 | |
| MaxCut | GSET (|V|>=3K) | Average Gap to Best Known Cut116.8 | 18 | |
| Maximum Independent Set | ER [9000-11000] | MIS Size373.3 | 13 | |
| Maximum Independent Set | MIS RB-small | Average Objective Value18.88 | 10 | |
| Maximum Cut | BA-[800-1200] (test) | Size2.96e+3 | 7 |