Abstract: Traditional deep learning methods are suboptimal in classifying ambiguity features, which often arise in noisy and hard to predict categories, especially, to distinguish semantic scoring.

Combinatorial optimization problems are often encountered in real-world applications, including logistics, scheduling and ...