Deep neural architectures for highly imbalanced data in bioinformatics

In this work, we present two new variants to the deepSOM model: the deep elastic SOM (deSOM) and the deep ensemble elastic SOM (deeSOM), which overcome the mentioned issues. In deSOM the number of deep levels not only grows automatically, but also the size of each layer is expanded adaptively accord...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Bugnon, Leandro A., Yones, Cristian, Milone, Diego H., Stegmayer, Georgina
Formato: Objeto de conferencia Resumen
Lenguaje:Inglés
Publicado: 2019
Materias:
Ciencias Informáticas
Bioinformatics
Pre-miRNA classification
Deep neural architectures
High class imbalance
Acceso en línea:http://sedici.unlp.edu.ar/handle/10915/87801
Aporte de:
SEDICI (UNLP) de Universidad Nacional de La Plata
Descripción
Sumario:In this work, we present two new variants to the deepSOM model: the deep elastic SOM (deSOM) and the deep ensemble elastic SOM (deeSOM), which overcome the mentioned issues. In deSOM the number of deep levels not only grows automatically, but also the size of each layer is expanded adaptively according to the data at each level, thus pre-miRNA neurons can be re-organized in a larger space.

Ejemplares similares