Name
Abstract | ||
|---|---|---|
Deep neural networks have achieved great success on a variety of machine learning tasks. There are many fundamental and open questions yet to be answered, however. We introduce the Extended Data Jacobian Matrix (EDJM) as an architecture-independent tool to analyze neural networks at the manifold of interest. The spectrum of the EDJM is found to be highly correlated with the complexity of the learned functions. After studying the effect of dropout, ensembles, and model distillation using EDJM, we propose a novel spectral regularization method, which improves network performance. |
Year | Venue | DocType |
|---|---|---|
2016 | ICML'16 Proceedings of the 33rd International Conference on International Conference on Machine Learning - Volume 48 | Conference |
Citations | PageRank | References |
2 | 0.38 | 0 |
Authors | ||
9 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Shengjie Wang | 1 | 5 | 3.11 |
| Abdel-rahman Mohamed | 2 | 3772 | 266.13 |
| Rich Caruana | 3 | 4503 | 655.71 |
| Jeff Bilmes | 4 | 3420 | 289.94 |
| Matthai Philipose | 5 | 2705 | 228.09 |
| Matthew Richardson | 6 | 4655 | 411.67 |
| Krzysztof Geras | 7 | 75 | 7.45 |
| Gregor Urban | 8 | 93 | 5.14 |
| Özlem Aslan | 9 | 36 | 2.50 |