GA Approach to Optimize Training Client Set in Federated Learning

Federated learning, where the distribution of distributed data is unknown, more difficult and costly to train a central model with than traditional machine learning. In this study, we propose Learning Genetic Algorithm, which enables faster training at lower cost by providing an appropriate client selection method. A can have its own communication depending on sharing preference, based result client’s local update, select combination clients each round genetic algorithm. round, combinations are evaluated anew, continually explored. To evaluate algorithm, image dataset costs in two ways conducted federated learning for classification model. Experiments showed that proposed algorithm find efficient accelerate