EMPR support vector selection via non-negative matrix factorization

GÜRVİT E.

International Conference on Numerical Analysis and Applied Mathematics, ICNAAM 2024, Heraklion, Greece, 11 - 17 September 2024, vol.3489, (Full Text) identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 3489
  • Doi Number: 10.1063/5.0328489
  • City: Heraklion
  • Country: Greece
  • Marmara University Affiliated: Yes

Abstract

Data storage plays a more and more increasing role Data plays a role in our progress and advancement. Therefore storage and processing of large datasets are to be more efficient. Multivariate modeling constitutes one of the most important aspects in this regard. It has a wide range of applications. Enhanced Multivariance Products Representation (EMPR) can be used to reduce dimension and feature extraction in high-dimensional data. EMPR decomposes high-dimensional data into lower-dimensional terms via decorrelation. To this end, EMPR uses support vectors through which a kind of weight is identified. Therefore, in this work, it is taken into account to determine these support vectors by factorizing the non-negative matrix [1].