Machine Learning

The aim of classification is the verify of the existence of differences between the classes as a function of the considered variables and the formulation of a model that is able to assign each sample to the class to which it belongs.

The machine learning methods (Machine Learning - ML) are used to perform predictions from a limited set of available data. ML algorithms learn a predictive function using data extracted according to a given probability distribution of the data from the data universe.

We can differentiate:

  • Supervised learning: the system is given as input a data set, said training set, containing both the input data and the output data. From this data set, the algorithm of supervised learning is trying to build a model that can predict the response values ​​for a new set of input data. The supervised learning includes two categories of algorithms, those of regression and classification
    • We apply several algorithms as AdaBoost, C4.5, SVM also for remote sensing with pixel, object and hybrid approaches.
  • Unsupervised learning: do not provide a set of data already structured to where to draw experiences. The task of the algorithm will instead just to find structures and patterns hidden within the data provided.
    • Our research promotes the boosting the performance by using GPU (CUDA enabled)