Overmind, A Collaborative Decentralized Machine Learning Framework, the Interpretation of Network Behaviour

Abstract

This paper is an extension of work originally presented in Overmind: A Collaborative Decentralized Machine Learning Framework, which focused on the presentation of the Overmind framework. Overmind is the conceptual design framework for a decentralized machine learning network containing the multiple agents which are collaboratively performing their tasks and objectives. This network of agents can be dynamically changed when the new features are discovered and introduced to the Overmind with significant changes to a system performance. The network behaviour of Overmind, as loosely described in a previous paper, is now presented with detail and interpretation. In this paper, Overmind has been deployed and applied to the multiple dataset which generates and creates the different outcomes and network pattern, then, its capabilities are tested by adding features to the network. The result shows four possible stages: 1) initial stage, 2) connected stage, 3) fully connected stage, and 3) isolated node. This paper also presents the future works as opportunities for improvement of the Overmind framework.