Automation of Information Exchange Between Neural Networks

In recent years, there have been developments in neural networks and machine learning. Numerous platforms and tools for modeling A.I. are published. Compatibility between these platforms and tools is not always smooth because there are no common standards. This article describes the creation of a prototype of protocol that combine multiple neural networks into a common system. This approach offers solutions to machine learning problems related to the processing of large amounts of data.A protocol that allows the automation of the exchange of matrices with weights and bias between neural networks is investigated. The protocol monitors factors such as type of architecture, completeness of training data, network specialization and retraining opportunities. Methods for decentralization of the system and stability of the protocol are considered.