Contents I. Introduction
Mobile edge computing has been introduced in various fields since it can be realized in the environments of real-time calculation by locating close to the data source [1]. For example, the object recognition [2] [3] for road management and vehicular networks for updating distributed models [4] are researched. The discriminator of the edge device needs to work for local optimized problems. Thus, the discriminator must be updated for solving the local problems. However, since the edge devices are worked in limited areas, these cannot obtain novel learning data by themselves. Therefore these discriminators need to update by using the knowledge of other devices. In the previous study, Federated Learning [5], which is the method to update distributed models using only weights of the Neural Network, is proposed. However, since this approach is using the knowledge or data of multiple devices in full area for composing robust model generating, local optimized model is not expected. Therefore, in order to update each model for local optimized problems, a novel method can be performed to share the knowledge of another device. Thus, in this paper, a method of updating models using weights of the Neural Networks by transferring and fitting is proposed. It is expected models can be updated for solving local optimized problems while sharing the knowledge of another model that are effective for learning. This paper discusses the details of the method for updating models by utilizing weights of another model. In the last of this paper, the effectiveness of the proposed method is verified using actual data.
