Hands-on Artificial Evolution Through Brain Programming

Abstract

This paper is about the evolution of a bio-inspired methodologyhttp://evovision.cicese.mx that mimics the cortical visual pathways. The methodology has been extensively tested on problems with different levels of complexity with outstanding results. After a review of the main works, the problem of classification of digitized art is introduced. An image database of five classes downloaded from the Kaggle web site is used as a benchmark for evolutionary learning. A comparison with convolutional neural network from scratch and the well-known AlexNet is provided to illustrate the quality of the proposal in comparison with the state-of-the-art.