Evaluación genética (que no genómica) mediante boosting

• J. Casellas ^[1]
  1. [1] Universitat Autònoma de Barcelona

     Universitat Autònoma de Barcelona

     Barcelona, España

     
• Localización: XV Jornadas sobre Producción Animal: 14 y 15 de mayo de 2013, Zaragoza / Jorge Hugo Calvo Lacosta (aut.), Isabel Casasús Pueyo (aut.), Margalida Joy Torrens (aut.), Javier Álvarez Rodríguez (aut.), Luis Varona Aguado (aut.), Begoña Panea Doblado (aut.), Carlos Calvete Margolles (aut.), Joaquín Balcells Teres (aut.), Vol. 2, 2013, ISBN 978-84-695-7684-7, págs. 496-498
• Idioma: español
• Títulos paralelos:
  • Genetic evaluation by boosting
• Enlaces
  • Texto Completo Libro  1 (pdf)   2 (pdf) 
• Resumen

  • This research focuses on the adaptation of the boosting algorithm to perform genetic evaluation of livestock populations. Boosting combines rough and moderately inaccurate predictors known as “weak learners” into a prediction rule with potentially greater predictive ability than that of any of the individual weak learners. This approach was evaluated on simulated data sets and compared with mixed linear animal models solved through a standard Bayesian approach. Prediction abilities of both boosting and BLUP approaches were similar, without a clear pattern favoring one or the other method. In a similar way, mean square differences between simulated and predicted breeding values were almost equal, and slight advantages were randomly obtained by the boosting or the BLUP approach depending on the analyzed data set. This suggested that boosting must be viewed as an appealing alternative when selection decisions must be taken on individuals without phenotypic data.