Pruning method for embedded network model

A network model, embedded technology, applied in the field of pruning, can solve problems such as poor practicability, and achieve the effect of small model, fast running speed, and wide application range

Inactive Publication Date: 2019-05-14
NORTHWESTERN POLYTECHNICAL UNIV
View PDF1 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the shortcomings of poor practicability of existing prunin

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
  • Pruning method for embedded network model
  • Pruning method for embedded network model
  • Pruning method for embedded network model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0025] refer to Figure 1-2 .

[0026] The core of this method is to sample a number of input and output of the convolutional layer, calculate the evaluation coefficient of each channel based on the above data based on the lasso algorithm, select unimportant channels, and cut off some convolution kernels corresponding to the channels. Taking the driver state detection model as an example, the specific steps of the pruning method for the embedded network model of the present invention are as follows:

[0027] Step 1. First, collect data, modify and train a new model. The state of the driver is divided into 7 states: looking left, looking right, looking down, closing eyes, making a phone call, yawning and driving normally. First mark the data, and then modify mobilenetSSD to output 7 states. Based on the model trained on the imagenet dataset, fine-tune training on new tasks to obtain a new model that can detect driving states.

[0028] Step 2: Establish a network model, perfo...

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 discloses a pruning method for an embedded network model. The pruning method is used for solving the technical problem that an existing pruning method is poor in practicability. According to the technical scheme, the method comprises the following steps: firstly, establishing a mobienet SSD network model, and carrying out a forward operation to obtain data required by pruning calculation; Channels which are not important to a convolution layer calculation result are selected through lasso regression, and channels which have relatively low influence on a summation result in the channels are selected through a lasso algorithm; The Mobienet resolves an original layer of convolution into a channel separation convolution layer and a point convolution layer, and an input channel ofthe channel separation convolution layer is equal to an output channel of the channel separation convolution layer. According to the method, the reconstruction error is reduced to serve as the core,the lasso is used for picking out unimportant channels in all convolution layers, then channel trimming is conducted on all the convolution layers according to the special structure of the mobienet, compression acceleration of the mobienet SSD is completed, and the practicability is good.

Description

technical field [0001] The invention relates to a pruning method, in particular to an embedded network model-oriented pruning method. Background technique [0002] With the development of deep convolutional network models in the field of artificial intelligence, the performance of large-scale network models is getting higher and higher. However, when deployed on some embedded devices, they are limited to limited computing resources and run too slowly. [0003] For the accelerated compression method of the neural network model, the key is to use an algorithm to pick out those unimportant nodes in the network, and then delete these redundant nodes to complete the compression acceleration of the network model. In the literature "Designing energyefficient convolutional neural networks using energy-awarepruning.arXiv preprint arXiv:1611.05128, 2016.", Yang et al. performed pruning based on the method of parameter values, cutting out those nodes with relatively low parameter value...

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
IPC IPC(8): G06N3/08G06N3/04
Inventor 袁媛王琦华路路
Owner NORTHWESTERN POLYTECHNICAL UNIV
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