Learning Diverse and Discriminative Representations via the Principle of Maximal Coding Rate Reduction
To learn intrinsic low-dimensional structures from high-dimensional data that most discriminate between classes, we propose the principle of Maximal Coding Rate Reduction ($\text{MCR}^2$), an information-theoretic measure that maximizes the coding rate difference between the whole dataset and the su...
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Artículo |
| Publicado: |
2020
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2006.08558 http://arxiv.org/abs/2006.08558 |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|