A Systematic Analysis of Federated Learning

Abstract

Deep Learning is an emerging area which is applicable to deal with massive and high multi-dimensional data in various E-Commerce platforms. This data is called as Big Data which is available in the complex and unstructured format, in terms of Volume, Velocity, Validity, Veracity, Variety and Variability. The data used in such e-platforms, is revealed at multiple places affecting the User's private data and security. To overcome the issue of user's private data security, a new dimension of Deep Learning Approach is been introduced, which is called as “Federated Learning (FL)”. Federated Learning provides Privacy Preservation and security to the User's data, thereby enhancing the accuracy of the system. This paper discusses about FL, various types of FL along with the implementation results of FL with Matrix Factorization (For accurate recommendations) on Movielens 1M dataset [15]. The paper concludes with various FL threats, and applications.