distributed learning proposal to improve industrial processes

Abstract

A distributed learning algorithm has been developed, focused on leveraging valuable information from industrial processes of various clients. This algorithm significantly improves the predictive capabilities of Machine Learning models by allowing access to a larger pool of training data. This is achieved by sharing the weights of the models among different participants, without the need to exchange the data itself, ensuring that each client maintains the privacy and security of their information. Thus, this approach not only optimizes the performance of the models individually but also enhances the overall level of artificial intelligence applied in the industrial sector.