Convolution kernel cutting method based on entropy importance criterion model

A convolution kernel and importance technology, which is applied in the field of convolution kernel clipping based on the entropy importance criterion model, can solve the problems of convolutional neural network model parameters, huge calculation amount, and inapplicability, so as to meet real-time performance and accuracy Requirements, the effect of achieving compression and acceleration

Active Publication Date: 2019-08-13
电科瑞达(成都)科技有限公司
View PDF8 Cites 10 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In view of the above-mentioned problems or deficiencies, in order to overcome the problem that the convolutional neural network model cannot be applied to scenarios with high real-time requirements due to the large amount of parameters and calculations

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Convolution kernel cutting method based on entropy importance criterion model
  • Convolution kernel cutting method based on entropy importance criterion model
  • Convolution kernel cutting method based on entropy importance criterion model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0025] In order to make the purpose of the present invention, technical solutions and advantages clearer, the present invention takes the Cifar10 data set as the target recognition task as an example, adopts the VGG16 and ResNet18 models as the model benchmarks, and further describes the present invention, wherein the structure of the VGG16 model is shown in the appendix Figure 4 , see the attached ResNet18 model structure Figure 5 .

[0026] The Cifar10 training sample is a 32×32 optical image, and only the Cifar10 data set is displayed. The image data display is shown in Figure 2.

[0027] (1) The VGG model is tested on Cifar10

[0028] It can be concluded from Table 1 that based on the VGG16 model, three methods were tested and comparative experiments were done.

[0029] Table 1 Comparison experiment of VGG16 on Cifar10 dataset

[0030] Model Acc(%) Parameter amount (M) FLOPS(M) Compression ratio acceleration rate VGG16 88.39 14.73 313 1x ...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention belongs to the technical field of neural networks, and relates to a convolution kernel cutting method based on an entropy importance criterion model. In order to enable a convolutional neural network model with large parameter quantity, large calculated quantity and excellent performance to meet a real-time requirement in practical application. The invention provides a convolutionalkernel cutting method based on an entropy importance criterion model. The convolution kernel with less information amount is cut by taking a mode of solving image entropy of an activation channel of each convolution layer as a criterion for evaluating the importance of the corresponding convolution kernel. Therefore,, a small model which is excellent in performance and small in parameter amount and calculation amount is obtained, and the small model has the performance advantage and can meet the real-time performance and precision requirements of practical scene application.

Description

technical field [0001] The invention belongs to the technical field of neural networks, and relates to a model convolution kernel clipping method based on an entropy importance criterion. Background technique [0002] In recent years, the convolutional neural network has developed very rapidly. With the continuous improvement of the theory and the support of modern large-scale computing platforms, the convolutional neural network has made great progress. It has applications in different fields, and has shown very good performance in different applications. Compared with the traditional feature extraction method, the convolutional neural network, as a hierarchical feature extractor, can extract more diverse and abstract features, making up for the lack of features manually extracted by traditional methods. The diversity and abstraction of the features extracted by the convolutional neural network enable the convolutional neural network to fit many applications and be used in...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/08G06N3/04
CPCG06N3/082G06N3/045
Inventor 闵锐蒋霆
Owner 电科瑞达(成都)科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap