ABSTRACT

One of the most common processes by which computing systems achieve the capability to perform sophisticated tasks is called machine learning, and it essentially consists in learning by examples. The description “classical machine learning” has originated in consequence to the increasing growth of a new branch of machine learning, deep learning, which has demonstrated impressive capabilities by enabling feature learning. Feature selection and extraction are key steps in classical machine learning applications. Feature extraction is particularly important in machine learning imaging applications, as data must be extracted from the images in order to be passed to a classifier. The process of feature selection should be robust to transformations in data such as preprocessing, should select the features which bring the information which is more relevant to the classification task and should not include features which are redundant.