Description

The traditional Machine Learning (ML) implementation method is based on a regime called offline learning or batch learning, where in large amounts of data are collected in advance and stored for training and verification purposes. This approach works well for applications where the input data stream is stable and regular, with minimal fluctuation or noise. This approach can however lead to a degradation in performance if the input changes in any way: for instance, if the mean or variance of the input data changes, or if any additional noise or other interference becomes present. Every time there are sufficiently drastic changes in the input data, the system must be taken offline, re-trained, and re-deployed. Unlike the above offline learning mechanism, humans can learn continuallyfrom experience in a regime referred to as online learning. Biologically inspired online learning techniques can be emulated, allowing ML systems to learn continually from a stream of data. The main purpose of this activity would be to identify potential satcom functionality and applications that could benefit from the use of continual learning methodologies and to explore, develop, and simulate different continual learning implementation techniques for the identified satcom applications.