| Resumen |
In the field of machine learning, there exist effective models that achieve remarkable results. However, it is criticized from the perspective of the end user that many successful models behave like “black boxes.” In the desire for achieving excellent performances, the models become more complicated and less explainable. In this context, a major challenge has emerged: it is necessary to provide the intelligent models and algorithms with the property to be explained. In this article, a new paradigm is proposed: the minimalist machine learning. The conceptual cornerstone of the new paradigm is the assumption that it is possible to make the representation of the data of any problem of classification of patterns reduced to the Cartesian plane. The proposal is to design algorithms that are capable of achieving intelligent pattern classification in two dimensions (the plane) effectively, regardless of the number of attributes that the patterns contain. |
