Super-network training method and device, electronic equipment and storage medium

A training method and a training device technology, applied in the field of deep learning, can solve problems such as limiting the representation ability of neural networks, and achieve the effect of improving the representation ability

Pending Publication Date: 2020-04-03
BEIJING XIAOMI INTELLIGENT TECH CO LTD
View PDF0 Cites 5 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] In the above neural network architecture search scheme, in the single-path neural sub-network trained for each sampling, in order to ensure the fai

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
  • Super-network training method and device, electronic equipment and storage medium
  • Super-network training method and device, electronic equipment and storage medium
  • Super-network training method and device, electronic equipment and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Example Embodiment

[0065] The exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the present disclosure. On the contrary, they are merely examples of devices consistent with some aspects of the present disclosure as detailed in the appended claims.

[0066] At present, in the existing neural network architecture search (NAS), a single-path neural sub-network is usually used to implement, that is, only one sub-module is sampled from each layer in the super network, and the sampled sub-modules are connected in series to form a single Path neural sub-network; then, share the parameters of each layer sub-module from the super netwo...

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 relates to a super-network training method and device, electronic equipment and a storage medium. The super-network training method comprises the steps: obtaining a multi-path neural sub-network, and carrying out the training of the multi-path neural sub-network, so as to update the weight parameters of all substructures; synchronizing the weight parameter of each substructure in themultipath neural sub-network to the initial super-network; and ending the training and obtaining a target super network when the super network converges. Thus, under the condition that the characterization capability of a single path is limited, the multi-path neural sub-network is utilized to train the super-network, and the characterization capability of the super-network can be improved.

Description

technical field [0001] The present disclosure relates to the technical field of deep learning, and in particular to a hypernetwork training method and device, electronic equipment, and storage media. Background technique [0002] At present, in the existing neural network architecture search (NAS), a single-path neural sub-network is usually used to implement, that is, only one sub-module is sampled from each layer of the super-network, and the sampled sub-modules are sequentially connected in series to form a single-path neural sub-network. path neural sub-network; then, share the parameters of each layer of sub-modules from the super-network; after that, perform single-step training on this single-path neural sub-network. After the training of the single-path neural sub-network is completed, the parameters of each sub-module are shared with the super-network. In this way, the steps of "sampling-sharing parameters-single-step training-updating parameters" are repeated unti...

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): G06N3/08
CPCG06N3/08G06N3/047G06N3/045G06N3/082G06N3/049G06N3/04
Inventor 初祥祥张勃许瑞军王斌
Owner BEIJING XIAOMI INTELLIGENT TECH CO LTD
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