Math @ Duke
|
Publications [#335541] of Jianfeng Lu
Papers Published
- Zhu, W; Qiu, Q; Wang, B; Lu, J; Sapiro, G; Daubechies, I, Stop memorizing: A data-dependent regularization framework for intrinsic
pattern learning, vol. 1 no. 3
(May, 2018),
pp. 476-496, Society for Industrial & Applied Mathematics (SIAM) [doi]
(last updated on 2024/09/17)
Abstract: Deep neural networks (DNNs) typically have enough capacity to fit random data
by brute force even when conventional data-dependent regularizations focusing
on the geometry of the features are imposed. We find out that the reason for
this is the inconsistency between the enforced geometry and the standard
softmax cross entropy loss. To resolve this, we propose a new framework for
data-dependent DNN regularization, the
Geometrically-Regularized-Self-Validating neural Networks (GRSVNet). During
training, the geometry enforced on one batch of features is simultaneously
validated on a separate batch using a validation loss consistent with the
geometry. We study a particular case of GRSVNet, the Orthogonal-Low-rank
Embedding (OLE)-GRSVNet, which is capable of producing highly discriminative
features residing in orthogonal low-rank subspaces. Numerical experiments show
that OLE-GRSVNet outperforms DNNs with conventional regularization when trained
on real data. More importantly, unlike conventional DNNs, OLE-GRSVNet refuses
to memorize random data or random labels, suggesting it only learns intrinsic
patterns by reducing the memorizing capacity of the baseline DNN.
|
|
dept@math.duke.edu
ph: 919.660.2800
fax: 919.660.2821
| |
Mathematics Department
Duke University, Box 90320
Durham, NC 27708-0320
|
|