The High-Performance Solution with Federated Learning in Supply Chain System
Haitao Liu, Bo Ye, Zhi Qin, Jing Zhang · 2022
Supply chain is an important commercial pattern which used in industry, agriculture and service widely. However different parties have different information even though they are all in the same supply chain. That would results in information silo and the extra increasing on cost. On one hand, information could not transfer to other parties for itself benefits or intellectual property. On the other hand, other parties want to obtain supplementary information in order to make equitable trade and lower trade risk. A skillful method would be designed and employed for solve above problems in this paper. Federated Learning technology was introduced in order to solve the puzzle between information privacy and information sharing. In the concretely implement process, Vertical Federated Learning (VFL) model was constructed and trained for resolving several problems specially because the overlap of parties are more but the features are small in supply chain forward. Gradient descent methods and loss computation ways were also used in training process in order to advance the performance. A series of experiments were used to evaluate VFL in supply chain. Experimental results revealed that VFL used in supply chain was effective.