A work on the research of the architecture of Convolutional Neural Networks and how to be able to make it possible on private computer at home.
In this work, we use a cartesian genetic algorithm to design Convolutional Neural Networks based on the paper: "A Genetic Programming Approach to Designing Convolutional Neural Network Architectures" [arXiv]. We combine this algorithm with a predictor to approximate the power of Convolutional Neural Networks like E2epp [ResearchGate]. Furthermore, We have also implemented an equation to allow us to take into account the size of the network to prioritize a network with less layers over a network with the same accuracy but more layers.