Federated Learning for Accurate Object Classification

nou Systems, Inc. proposes utilizing the Artificial Intelligence (AI techniques of Vertical Federated Learning and Split Learning to improve object classification This training paradigm will also accommodate multiple sources without duplicating data. A Federated Neural Network (FNN) or split learning AI approach would allow data from disparate sensors to train on their own observations and proved parameters and weights to a trusted third-party to do classification. FNN's are very robust to non-IID (Identically and Dependently Distributed) data, which is inherent to using multiple sensors placed at different locations, viewing different aspects and stages of flight, and trained on different threat databases. The technique will aid in better identifying threats not in an a-priori database. Approved for Public Release | 21-MDA-11013 (19 Nov 21)