| （英） |
In federated learning, a method of distributed learning, a global model is constructed from only the training results without disclosing the training data of participating clients. By using the constructed global model, participating clients can perform highly accurate learning using a large amount of data held by other clients. Currently, federated learning is used in fields where privacy protection and security measures are important, such as pharmaceuticals and finance. In the future, it is expected to be used in applications for mobile terminals. However, in actual applications, multiple clients have different amount of data. In this study, we consider a fishing prediction using federated learning, and evaluate the impact of this system on fishing prediction through experiments. First, we construct a simple federated learning system in a virtual environment and investigate the effects of the number of clients and the amount of data on learning accuracy. |