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

Data processing method and device

A data processing and training method technology, applied in the field of machine learning, can solve problems such as adverse effects of learning, achieve accurate data processing results, accelerate training, and improve similarity effects

Active Publication Date: 2021-10-08
MEGVII BEIJINGTECH CO LTD
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Among the various methods of model search, weight sharing is a method that saves computing resources, such as non-differentiable models such as SuperNet and One-shot, and differentiable models such as DARTS and ProxylessNAS. Search algorithms all use this method, but the disadvantage of this method is that different operation operators of each layer of the network may output features with different distributions, which has an adverse effect on the learning of subsequent layers of the network

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
  • Data processing method and device
  • Data processing method and device
  • Data processing method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0019] The principle and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present invention, rather than to limit the scope of the present invention in any way.

[0020] like figure 1 As shown, an embodiment of the present invention proposes a method 100 for training a network to be searched. Wherein, the network to be searched may include a convolutional neural network. The method 100 includes steps S101-S105. The method 100 can be applied to various weight sharing based model search methods.

[0021] In some embodiments, on the basis of non-differentiable model search algorithms such as SuperNet and One-shot and differentiable model search algorithms such as DARTS and ProxylessNAS, a teacher network is added to guide the search The training process of the model. In some embodiments...

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 provides a data processing method and device. The data processing method includes: a step of training the network to be searched, training the network to be searched according to a training method for the network to be searched until the network to be searched converges to obtain a trained network to be searched; wherein the network to be searched includes A convolutional neural network; a model-specific search step, a model-specific training step, and a data processing step. Wherein, the training method of the network to be searched includes the step of acquiring network loss, acquiring the feature of the middle layer, matching the feature of the middle layer, calculating the comprehensive loss and feeding back the comprehensive loss. The invention can speed up the training of the model and facilitate the model search.

Description

technical field [0001] The present invention generally relates to the field of machine learning, and more specifically relates to a data processing method and device. Background technique [0002] In recent years, deep learning has achieved end-to-end feature extraction, which is a huge improvement compared to manual feature extraction, which has made great progress in tasks such as computer vision, natural language processing, and speech recognition; and often a better neural network The emergence of network architecture means that a certain degree of performance improvement can be obtained on various tasks. But the neural network architecture relies heavily on manual design, which is very time-consuming and energy-consuming even for an experienced machine learning practitioner. [0003] Therefore, several approaches to model search have recently emerged to automate the design of neural network architectures and represent the future direction of machine learning. Among th...

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 Patents(China)
IPC IPC(8): G06N3/04G06K9/62
CPCG06N3/045G06F18/214
Inventor 陈程鹏
Owner MEGVII BEIJINGTECH 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