English
 
User Manual Privacy Policy Disclaimer Contact us
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Paper

On Implicit Filter Level Sparsity in Convolutional Neural Networks

MPS-Authors
/persons/resource/persons129023

Mehta,  Dushyant
Computer Graphics, MPI for Informatics, Max Planck Society;

/persons/resource/persons45610

Theobalt,  Christian
Computer Graphics, MPI for Informatics, Max Planck Society;

Locator
There are no locators available
Fulltext (public)

arXiv:1811.12495.pdf
(Preprint), 8MB

Supplementary Material (public)
There is no public supplementary material available
Citation

Mehta, D., Kim, K. I., & Theobalt, C. (2018). On Implicit Filter Level Sparsity in Convolutional Neural Networks. Retrieved from http://arxiv.org/abs/1811.12495.


Cite as: http://hdl.handle.net/21.11116/0000-0002-E920-3
Abstract
We investigate filter level sparsity that emerges in convolutional neural networks (CNNs) which employ Batch Normalization and ReLU activation, and are trained with adaptive gradient descent techniques and L2 regularization (or weight decay). We conduct an extensive experimental study casting these initial findings into hypotheses and conclusions about the mechanisms underlying the emergent filter level sparsity. This study allows new insight into the performance gap obeserved between adapative and non-adaptive gradient descent methods in practice. Further, analysis of the effect of training strategies and hyperparameters on the sparsity leads to practical suggestions in designing CNN training strategies enabling us to explore the tradeoffs between feature selectivity, network capacity, and generalization performance. Lastly, we show that the implicit sparsity can be harnessed for neural network speedup at par or better than explicit sparsification / pruning approaches, without needing any modifications to the typical training pipeline.