File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Diverse neural network learns true target functions

TitleDiverse neural network learns true target functions
Authors
Issue Date2017
Citation
Proceedings of the 20th International Conference on Artificial Intelligence and Statistics, AISTATS 2017, 2017 How to Cite?
AbstractNeural networks are a powerful class of functions that can be trained with simple gradient descent to achieve state-of-the-art performance on a variety of applications. Despite their practical success, there is a paucity of results that provide theoretical guarantees on why they are so effective. Lying in the center of the problem is the difficulty of analyzing the non-convex loss function with potentially numerous local minima and saddle points. Can neural networks corresponding to the stationary points of the loss function learn the true target function? If yes, what are the key factors contributing to such nice optimization properties? In this paper, we answer these questions by analyzing one-hidden-layer neural networks with ReLU activation, and show that despite the non-convexity, neural networks with diverse units have no spurious local minima. We bypass the non-convexity issue by directly analyzing the first order optimality condition, and show that the loss can be made arbitrarily small if the minimum singular value of the “extended feature matrix” is large enough. We make novel use of techniques from kernel methods and geometric discrepancy, and identify a new relation linking the smallest singular value to the spectrum of a kernel function associated with the activation function and to the diversity of the units. Our results also suggest a novel regularization function to promote unit diversity for potentially better generalization.
Persistent Identifierhttp://hdl.handle.net/10722/341271

 

DC FieldValueLanguage
dc.contributor.authorXie, Bo-
dc.contributor.authorLiang, Yingyu-
dc.contributor.authorSong, Le-
dc.date.accessioned2024-03-13T08:41:30Z-
dc.date.available2024-03-13T08:41:30Z-
dc.date.issued2017-
dc.identifier.citationProceedings of the 20th International Conference on Artificial Intelligence and Statistics, AISTATS 2017, 2017-
dc.identifier.urihttp://hdl.handle.net/10722/341271-
dc.description.abstractNeural networks are a powerful class of functions that can be trained with simple gradient descent to achieve state-of-the-art performance on a variety of applications. Despite their practical success, there is a paucity of results that provide theoretical guarantees on why they are so effective. Lying in the center of the problem is the difficulty of analyzing the non-convex loss function with potentially numerous local minima and saddle points. Can neural networks corresponding to the stationary points of the loss function learn the true target function? If yes, what are the key factors contributing to such nice optimization properties? In this paper, we answer these questions by analyzing one-hidden-layer neural networks with ReLU activation, and show that despite the non-convexity, neural networks with diverse units have no spurious local minima. We bypass the non-convexity issue by directly analyzing the first order optimality condition, and show that the loss can be made arbitrarily small if the minimum singular value of the “extended feature matrix” is large enough. We make novel use of techniques from kernel methods and geometric discrepancy, and identify a new relation linking the smallest singular value to the spectrum of a kernel function associated with the activation function and to the diversity of the units. Our results also suggest a novel regularization function to promote unit diversity for potentially better generalization.-
dc.languageeng-
dc.relation.ispartofProceedings of the 20th International Conference on Artificial Intelligence and Statistics, AISTATS 2017-
dc.titleDiverse neural network learns true target functions-
dc.typeConference_Paper-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.scopuseid_2-s2.0-85083936469-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats