DiffUCO: A Diffusion Mannequin Framework for Unsupervised Neural Combinatorial Optimization
Sampling from complicated, high-dimensional goal distributions, such because the Boltzmann distribution, is essential in lots of scientific fields. For example, predicting molecular configurations depends upon one of these sampling. Combinatorial Optimization (CO) could be seen as a distribution studying drawback the place the samples correspond to options of CO issues, however it’s difficult to attain…
