Accelerating Diffusion-based Combinatorial Optimization Solvers by Progressive Distillation
This work addresses the slow inference problem for researchers and practitioners using diffusion models for NP-complete combinatorial optimization, representing an incremental improvement.
The paper tackles the inefficiency of diffusion-based combinatorial optimization solvers by applying progressive distillation to reduce inference steps, achieving a 16x speedup with only 0.019% performance degradation on the TSP-50 dataset.
Graph-based diffusion models have shown promising results in terms of generating high-quality solutions to NP-complete (NPC) combinatorial optimization (CO) problems. However, those models are often inefficient in inference, due to the iterative evaluation nature of the denoising diffusion process. This paper proposes to use progressive distillation to speed up the inference by taking fewer steps (e.g., forecasting two steps ahead within a single step) during the denoising process. Our experimental results show that the progressively distilled model can perform inference 16 times faster with only 0.019% degradation in performance on the TSP-50 dataset.