Advancing Machine Learning with Distributed and Edge Computing

Abstract

The amount of data used by modern machine learning algorithms is increasing, which presents several challenges. First and foremost, the data does not come from a single repository, but is distributed across multiple sources, often in different geographic locations. Another challenge is the high hardware requirements needed to process the data, which can be prohibitively expensive. The proposed approach allows a classifier to run in a distributed environment, simulating a real-world scenario where each node is as close to the data as possible, eliminating the need for a single data source. Furthermore, distributing the computational load of classification across multiple machines can be beneficial in an Internet of Things environment, avoiding the need to purchase expensive equipment.