A neural network for handwritten pattern recognition

Neuroscientists are mainly interested in the study of the behavior of a whole set of cells, while AI researchers study how to extract, freeze and analyze the process, by means of models, and learn the specificity of each kind of neuron´s behavior. Our aim is to describe a neural network based system, inspired by the optical structure of human cortex area and its application to pattern recognition. We adopted an unsupervised learning algorithm that is applied to each presented pattern. The system´s most relevant feature is its insensibility to spatial shift variance. The system´s behavior is based on the identification and decomposition of the relevant input image characteristics, that will be used to achieve classification. Finally, we briefly describe a case study followed by some conclusion