Reproducing expirements

In the final checkpoint of my work, the general experiments given in the article were performed, as well as additional experiments related to the implementation of the Dmeta algorithm, which were presented in the article. You can also view the results and the final presentation in the reports folder.

UGA & FedMeta implementation

During the second three weeks, the main methods of distributed optimization given in the article were implemented. The algorithms were tested on the basis of a distributed client system, which was trained on the FEMINIST dataset, solving the problem in a convolutional way.

All the algorithms considered in the article showed an increase in performance, which will have a positive impact on further analysis and operation.

Problem statement

During the first three weeks, an article about optimization methods in Federated Learning was selected and studied. The thesis and significance of the problems indicated by the author of the article were studied.