Authors: Guangqun Chen
In this paper a new type of layer named Structure Composing Layer (SCLayer) aims to reduce parameters' numbers and decrease computation cost is proposed. SCLayer combines simple structures into complex structure with less parameters. Compared to convolution layer with filter kernel size 3x3, SCLayer reduces parameters' numbers to 1/9 of convolution layer's. Based on VGG-16 with SCLayers a neural network, named Snowflake Net (SfNet), is designed and evaluated on dataset CIFAR-10 and CALTECH-256. Experiment results show that comparing with VGG-16 training time reduces more than 12.9%, classification accuracy increases about 0.18% on CIFAR-10, classification accuracy increases about 3.11% on CALTECH-256. During training process nonlinear phenomenon Accuracy Valley is observed. Influences of nonlinear phenomena on models' generalization are discussed.
Comments: 18 Pages.
Unique-IP document downloads: 20 times
Vixra.org is a pre-print repository rather than a journal. Articles hosted may not yet have been verified by peer-review and should be treated as preliminary. In particular, anything that appears to include financial or legal advice or proposed medical treatments should be treated with due caution. Vixra.org will not be responsible for any consequences of actions that result from any form of use of any documents on this website.
Add your own feedback and questions here:
You are equally welcome to be positive or negative about any paper but please be polite. If you are being critical you must mention at least one specific error, otherwise your comment will be deleted as unhelpful.