We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Computer Science > Machine Learning

Title: Optimization Landscape and Expressivity of Deep CNNs

Abstract: We analyze the loss landscape and expressiveness of practical deep convolutional neural networks (CNNs) with shared weights and max pooling layers. We show that such CNNs produce linearly independent features at a "wide" layer which has more neurons than the number of training samples. This condition holds e.g. for the VGG network. Furthermore, we provide for such wide CNNs necessary and sufficient conditions for global minima with zero training error. For the case where the wide layer is followed by a fully connected layer we show that almost every critical point of the empirical loss is a global minimum with zero training error. Our analysis suggests that both depth and width are very important in deep learning. While depth brings more representational power and allows the network to learn high level features, width smoothes the optimization landscape of the loss function in the sense that a sufficiently wide network has a well-behaved loss surface with almost no bad local minima.
Comments: Accepted at ICML 2018
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Optimization and Control (math.OC); Machine Learning (stat.ML)
Cite as: arXiv:1710.10928 [cs.LG]
  (or arXiv:1710.10928v2 [cs.LG] for this version)

Submission history

From: Quynh Nguyen [view email]
[v1] Mon, 30 Oct 2017 13:24:28 GMT (89kb,D)
[v2] Wed, 6 Jun 2018 12:49:58 GMT (99kb,D)

Link back to: arXiv, form interface, contact.