Paper Info
Title
Optimization on Product Submanifolds of Convolution Kernels.
Abstract
We address a problem of optimization on product of embedded submanifolds of convolution kernels (PEMs) in convolutional neural networks (CNNs). First, we explore metric and curvature properties of PEMs in terms of component manifolds. Next, we propose a SGD method, called C-SGD, by generalizing the SGD methods employed on kernel submanifolds for optimization on product of different collections of kernel submanifolds. Then, we analyze convergence properties of the C-SGD considering sectional curvature properties of PEMs. In the theoretical results, we expound the constraints that can be used to compute adaptive step sizes of the C-SGD in order to assure the asymptotic convergence.
Year
Venue
Field
2017
arXiv: Computer Vision and Pattern Recognition
Convergence (routing),Curvature,Normalization (statistics),Pattern recognition,Computer science,Convolutional neural network,Convolution,Artificial intelligence,Joint spaces,Computation
DocType
Volume
Citations 
Journal
abs/1701.06123
0
PageRank 
References 
Authors
0.34
2
2
Name
Order
Citations
PageRank
Mete Ozay110614.50
Takayuki Okatani249250.10