Uma abordagem para orquestração de funções de aprendizado de máquina virtualizadas em containers

Abstract

Machine learning is one of the fastest growing areas in computing, one of the factors behind this growth is that smart application utilization is becoming increasingly common today. Along with the increased computing power of the devices around us, it has been possible to develop new applications capable of achieving better results in the creation and execution of artificial intelligence. Using distributed systems we are able to separate the functions related to learning from machine for best performance. However, unconsciousness may occur in the distributed execution of certain tasks as functions require specific libraries to be performed. One approach that solves this type of problem is virtual machines and containers, which can be run as instantiation services using part of user machine processing to virtualize a process. This paper presents an approach to virtualizing machine learning functions using container dockers, making use of a distributed system at the network edge for processing and considering the capacity of each node of the system.