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

Method, device and apparatus for determining structure of neural network,and readable medium

A network structure and neural network technology, applied in the field of computer vision, can solve the problems of poor neural network processing effect and inappropriate data processing.

Active Publication Date: 2019-02-19
BEIJING BYTEDANCE NETWORK TECH CO LTD
View PDF11 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In the prior art, most of the relatively mature neural networks such as RCNN, TOLO, and SSD are used for image processing. However, the inventors found during the research process that these mature neural networks are not suitable for all data processing. For example, when processing some These neural networks are less effective when processing images

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
  • Method, device and apparatus for determining structure of neural network,and readable medium
  • Method, device and apparatus for determining structure of neural network,and readable medium
  • Method, device and apparatus for determining structure of neural network,and readable medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0032] figure 1 It is a flowchart of a method for determining the structure of a neural network provided by Embodiment 1 of the present disclosure. This embodiment is applicable to the situation of determining the structure of a neural network. The method can be performed by a device for determining the structure of a neural network. The device can It consists of hardware and / or software, and is integrated into an electronic device, which may be a server or a terminal. to combine figure 1 , the method provided by the embodiment of the present disclosure specifically includes the following operations:

[0033] S110. Sample the current network structure by using a sampler. Continue to execute S120.

[0034] The sampler is used to sample the network structure of the neural network. Each run of the sampler can sample at least one current network structure.

[0035] S120. Calculate an objective function value of the current network structure. Continue to execute S130.

[003...

Embodiment 2

[0047] figure 2 It is a flowchart of a method for determining the structure of a neural network provided in Embodiment 2 of the present disclosure. This embodiment further optimizes the optional implementations of the above embodiments. Optionally, the operation "calculate the objective function value of the current network structure" is refined into "calculate the accuracy rate and / or running time of the current network structure; according to The accuracy rate and / or running time of the current network structure, and obtain the objective function value", so as to sample a network structure with a higher accuracy rate or a shorter running time. Optionally, after the operation "adjust the parameters of the sampler according to the objective function value", add "initialize the network parameters of the current network structure; calculate the network parameters of the current network structure through the data set" to obtain not only the network structure, but also the approp...

Embodiment 3

[0071] Figure 3a It is a flowchart of a method for determining the structure of a neural network provided in Embodiment 3 of the present disclosure. This embodiment further optimizes the optional implementations of the above-mentioned embodiments. Optionally, "sampling the current network structure through the sampler" is optimized as "inputting the predefined network layer code into the sampler to obtain the current Network structure coding; according to the current network structure coding, construct the current network structure", which provides the sampling method of the network structure. combine Figure 3a , the method provided in this embodiment specifically includes the following operations:

[0072] S310. Input the predefined network layer codes into the sampler to obtain the current network structure codes.

[0073] A network unit includes at least one network layer, such as a convolutional layer, a pooling layer, and a connection layer. If the stacking times of...

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

Embodiments of the present disclosure disclose a method, device and apparatus for determining the structure of a neural network, and readable medium. The method comprises the following steps: samplingthe current network structure through a sampler; calculating an objective function value of the current network structure; adjusting a parameter of the sampler according to the objective function value; returning to the operation of performing sampling out of the current network structure by the sampler until the objective function value reaches the preset function value and / or the number of adjustments reaches the number of times threshold. The disclosed embodiment can realize automatic searching of the network structure, and is not confined to a fixed network structure, but continuously obtains a new and higher quality network structure through the value of the objective function, and can be applied to data processing of almost all kinds of scenes.

Description

technical field [0001] Embodiments of the present disclosure relate to computer vision technology, and in particular to a method, device, device, and readable medium for determining the structure of a neural network. Background technique [0002] With the development of computer vision, data such as images and sounds can be processed through neural networks, such as object detection, object tracking, segmentation, and classification of objects in images. [0003] With the improvement of user needs and the development of terminal technology, higher requirements are put forward for the accuracy and speed of data processing, which requires a neural network with better processing effect. In the prior art, most of the relatively mature neural networks such as RCNN, TOLO, and SSD are used for image processing. However, the inventors found during the research process that these mature neural networks are not suitable for all data processing. For example, when processing some These...

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): G06N3/04
CPCG06N3/04
Inventor 胡耀全
Owner BEIJING BYTEDANCE NETWORK TECH CO LTD
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

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com