Estudio del comportamiento de métodos basados prototipos y en relaciones de similitud ante “hubness”

• Yanela Rodríguez Alvarez ^[1] ; Rafael Bello Pérez ^[2] ; Yailé Caballero Mota ^[1] ; Yaima Filiberto Cabrera ^[1] ; Yumilka Fernández Hernández ^[1] ; Mabel Frías Hernández ^[1]
  1. [1] Universidad de Camagüey

     Universidad de Camagüey

     Cuba

     
  2. [2] Universidad Central de Las Villas

     Universidad Central de Las Villas

     Cuba

     
• Localización: Revista Cubana de Ciencias Informáticas, ISSN-e 2227-1899, Vol. 11, Nº. 2, 2017
• Idioma: español
• Títulos paralelos:
  • A study of the behavior of methods based on prototypes and similarity relations in the face of “hubness”
• Enlaces
  • Texto completo
• Dialnet Métricas: 1 Cita
• Resumen
  • español

    El fenómeno hubness, es un aspecto del curso de la dimensionalidad descrito recientemente, que está relacionado con la disminución de la especificidad de las similitudes entre los puntos en un espacio de alta dimensión; lo cual va en detrimento de los métodos de aprendizaje automático. En este trabajo se evalúa el impacto del fenómeno hubness en la clasificación utilizando un enfoque basado en prototipos. El estudio experimental realizado demuestra que los métodos de generación y selección de prototipos estudiados ofrecen resultados comparables contra otros métodos basados en el enfoque kNN, encontrados en la literatura, los cuales son hubness-consientes y están diseñados específicamente para lidiar con este problema. Teniendo en cuenta los resultados alentadores de este estudio y las bondades de los métodos basados en prototipos es posible asegurar que la utilización de los mismos permitirá mejorar el desempeño de los sistemas que manejen datos de altas dimensiones y bajo la asunción de hubness.

  • English

    The hubness phenomenon, is an aspect of the curse of dimensionality recently described, that is related to the diminishment of specificity in similarities between points in a high-dimensional space; which is detrimental to the machine learning methods. This paper deals with evaluating the impact of hubness phenomenon on classification based on the nearest prototype. Experimental results show that the studied methods of generation and selection of prototypes offer comparable results against others methods based on kNN approach, found in the literature, which are hubness aware and are specifically designed to deal with this problem. Based on these encouraging results and the extensibility of methods based on prototypes, it is possible argue that it might be beneficial to use them in order to improve system performance in high dimensional data under the assumption of hubness.

• Referencias bibliográficas
  • Alcalá, J,Fernández, A,Luengo, J,Derrac, J,García, S,Sánchez, L,Herrera, F. Keel data-mining software tool: Data set repository, integration...
  • Bello-García, M,García-Lorenzo, M. M,Bello, R. (2012). A method for building prototypes in the nearest prototype approach based on similarity...
  • Bezdek, J. C,Kuncheva, L. I. Nearest prototype classifier designs: An experimental study. 16. 1445-1473
  • Bolei, Z,Khosla, A,Lapedriza, A,Oliva, A,Torralba, A. (2015). Object detectors emerge in Deep Scene CNNs..
  • Choi, W,Pantofaru, C,Savarese, S. A general framework for tracking multiple people from a moving camera.. 35. 1577-1591
  • Demšar, J. (2006). Statistical comparisons of classifiers over multiple data sets.. The Journal of Machine Learning Research.
  • Duda, R,Hart, P,Stork, D. (2001). Pattern classification. 1. 335-339
  • Fernández Hernández, Yumilka,Bello, B,Filiberto, Y,Frías, M,Coello Blanco, L,Caballero, Y. An Approach for Prototype Generation based on Similarity...
  • Filiberto, Y,Bello, R,Caballero, Y,Larrua, R. (2010). Using PSO and RST to predict the resistant capacity of connections in composite structures...
  • Filiberto, Y,Caballero, Y,Larrua, R,Bello, R. (2010). A method to build similarity relations into extended rough set theory.
  • Frias, M,Filiberto, Y,Fernández, Y,Caballero, Y,Bello, R. (2015). Prototypes selection based on similarity relations for classification problems...
  • Gao, L,Song, J,Liu, X,Shao, J,Liu, J,Shao, J. (2015). Learning in high-dimensional multimedia data: the state of the art. 1
  • Gao, L,Song, J,Nie, F,Yan, Y,Sebe, N,Tao Shen, H. (2015). Optimal Graph Learning with Partial Tags and Multiple Features for Image and Video...
  • Garcia, S,Derrac, J,Cano, J. R.,Herrera, F. Prototype selection for nearest neighbor classification: Taxonomy and empirical study. IEEE Transactions....
  • García, S,Fernández, A,Luengo, J,Herrera, F. Advanced nonparametric tests for multiple comparisons in the design of experiments in computational...
  • García, S,Herrera, F. Evolutionary undersampling for classification with imbalanced datasets: Proposals and taxonomy. 17. 275-306
  • Jin, X.-B,Liu, C.-L,Hou, X. Regularized margin-based conditional log-likelihood loss for prototype learning. 43. 2428-2438
  • Keller, J. M,Gray, M. R,Givens, J. A. (1985). A fuzzy k-nearest neighbor algorithm.. IEEE.
  • Khosla, A,An, B,Lim, J. J,Torralba, A. (2014). Looking beyond the visible scene. Paper presented at the Computer Vision and Pattern Recognition...
  • Low, T,Borgelt, C,Stober, S,Nürnberger, A. (2013). The Hubness Phenomenon: Fact or Artifact?. Springer.
  • Nguyen, A,Yosinski, J,Clune, J. (2015). Deep neural networks are easily fooled: High confidence predictions for unrecognizable images. IEEE...
  • Radovanović, M,Nanopoulos, A,Ivanović, M. (2009). Nearest neighbors in high-dimensional data: The emergence and influence of hubs..
  • Tomašev, N,Mladenić, D. (2012). Nearest neighbor voting in high dimensional data:: Learning from past occurrences. Computer Science and Information...
  • Tomašev, N,Mladenić, D. Hubness-aware shared neighbor distances for high-dimensional k-nearest neighbor classification. 39. 89-122
  • Tomasev, N,Radovanović, M,Mladenić, D,Ivanović, M. (2011). A probabilistic approach to nearest-neighbor classification: naive hubness bayesian...
  • Tomašev, N,Radovanović, M,Mladenić, D,Ivanović, M. Hubness-based fuzzy measures for high-dimensional k-nearest neighbor classification. 5....
  • Triguero, I,Derrac, J,Garcia, S,Herrera, F. A taxonomy and experimental study on prototype generation for nearest neighbor classification....
  • Weinberger, K. Q,Saul, L. K. (2009). Distance metric learning for large margin nearest neighbor classification. The Journal of Machine Learning...
  • Wiskott, L. How to solve classification and regression problems on high-dimensional data with a supervised extension of slow feature analysis.....
  • Witten, I. H,Frank, E. (2005). Data Mining: Practical machine learning tools and techniques. Morgan Kaufmann..
  • Yao, B,Khosla, A,Fei-Fei, L. (2011). Classifying actions and measuring action similarity by modeling the mutual context of objects and human...
  • Yao, Y. (2000). Granular computing: basic issues and possible solutions.
  • Zhou, X,Chen, L,Zhang, Y,Cao, L,G, Huang,Wang, C. (2015). Online video recommendation in sharing community.