Error bounds for deep ReLU networks using the Kolmogorov--Arnold superposition theorem
We prove a theorem concerning the approximation of multivariate functions by deep ReLU networks, for which the curse of the dimensionality is lessened. Our theorem is based on a constructive proof of the Kolmogorov--Arnold superposition theorem, and on a subset of multivariate continuous functions w...
Kaydedildi:
| Asıl Yazarlar: | , |
|---|---|
| Materyal Türü: | Makale |
| Baskı/Yayın Bilgisi: |
2019
|
| Konular: | |
| Online Erişim: | https://arxiv.org/abs/1906.11945 http://arxiv.org/abs/1906.11945 |
| Etiketler: |
Etiket eklenmemiş, İlk siz ekleyin!
|