Incremental Learning via Rate Reduction
Current deep learning architectures suffer from catastrophic forgetting, a failure to retain knowledge of previously learned classes when incrementally trained on new classes. The fundamental roadblock faced by deep learning methods is that deep learning models are optimized as "black boxes," making...
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Artículo |
| Publicado: |
2020
|
| Materias: | |
| Acceso en línea: | https://arxiv.org/abs/2011.14593 http://arxiv.org/abs/2011.14593 |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|