Título: A quantum inspired methodology for enhancement of data discrimination power
Autores: Souza, Rosilda B.; Pereira, Emeson J. S.; Ferreira, Tiago A. E.
Resumo: The proposed methodology is a process for enhancement the data power discrimination based on the Cover theorem with a quantum inspiration. Given a set of data with several class, the proposed process consists in increasing its dimension in order to try become the non linearly separable problem into a linearly separable problem. Also is supposed that the data are observables in the quantum world, i.e., the data (real number) are expected value measurements of the given transformation with respect to a quantum state (complex numbers). Therefore, the methodology applies a Genetic Algorithm for search the inverse mapping of the expected value measurements, transforming the real number into complex number, subject to the constraint of magnitude conservation. The traditional methods of classification like K-means, KNN and LDA were applied to benchmark classification problems in two conditions: raw data set and transformed data set with the proposed methodology. The comparison of the classification results are presented, indicating a enhancement in the data power discrimination when the proposed pre-processing is applied.
Código DOI: 10.21528/CBIC2013-141
Artigo em pdf: bricsccicbic2013_submission_141.pdf
Arquivo BibTex: bricsccicbic2013_submission_141.bib