Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Neural network decoupling method

A neural network and loss function technology, applied in the field of neural network decoupling, can solve problems such as poor interpretability, achieve the effects of enhancing robustness, increasing computational load, and reducing computational load

Pending Publication Date: 2020-09-29
南强智视(厦门)科技有限公司
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Based on the above analysis, aiming at the disadvantage of poor interpretability of neural network, the inventor proposed an interpretable neural network decoupling method in consideration of the characteristics of neural network itself, and this case arose from this

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
  • Neural network decoupling method
  • Neural network decoupling method
  • Neural network decoupling method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0066] The technical solutions and beneficial effects of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0067] Such as figure 1 As shown, the present invention provides a neural network decoupling method, comprising the following steps:

[0068] Step 1. Construct a structural control module in each convolutional layer of the neural network to select the filters involved in the calculation;

[0069] For the modeling of the calculation path, it is first necessary to be able to automatically select the filters that need to participate in the calculation based on the input. Therefore, the core problem is to build a structural control module G, whose input is the input X of this layer, and the output is which filters need to be selected. The device participates in the calculation of Z, which can also be called structural coding. For the structural control module G, the input X is firstly pooled globally to obtain its corre...

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 neural network decoupling method, which comprises the following steps of: 1, for each convolutional layer of a neural network, selecting a filter participating in calculationaccording to the importance of each filter in the convolutional layer; and 2, training the neural network obtained in the step 1 based on a mutual information loss function, a KL-divergence loss function and a sparse loss function. According to the method, the structure of the neural network is decoupled, so that different calculation paths can be generated for different inputs, the working principle of the neural network is further explained, the inference of the neural network is accelerated, and the robustness of the neural network is enhanced.

Description

technical field [0001] The invention relates to an interpretable neural network decoupling method. Background technique [0002] In recent years, with the rapid development of hardware GPU and the advent of the era of big data, deep learning has developed rapidly and has swept all fields of artificial intelligence, including speech recognition, image recognition, video tracking, natural speech processing, etc. , Video field. Deep learning technology breaks through traditional technical methods and greatly improves the recognition performance in various fields, especially the powerful self-feature representation ability of convolutional neural networks (CNNs), which makes it widely used in image recognition[1-4] , target detection [5-7], image retrieval [8] and other fields. However, due to the high storage and inexplicability of the convolutional neural network model, the model cannot be directly embedded into mobile devices with limited storage space, and it cannot be app...

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): G06N3/04G06N3/08
CPCG06N3/08G06N3/048G06N3/045
Inventor 王振宁许金泉王溢曾尔曼
Owner 南强智视(厦门)科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products