Unlock instant, AI-driven research and patent intelligence for your innovation.

Search method

A search method and network search technology, applied in the direction of neural learning methods, other database retrieval, special data processing applications, etc., can solve the problem of consuming computing resources and achieve multi-value effects

Active Publication Date: 2020-05-15
SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
View PDF3 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

At present, more mature solutions generally include NAS (Neural Architecture Searching) methods, the self-optimization platform AotuML provided by Google, etc., but still need to consume a lot of computing resources

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
  • Search method
  • Search method
  • Search method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0039] Hereinafter, specific embodiments of the present application will be described in detail with reference to the accompanying drawings. According to these detailed descriptions, those skilled in the art can clearly understand the present application and can implement the present application. Without departing from the principle of the present application, the features in different embodiments can be combined to obtain new implementations, or some features in certain embodiments can be replaced to obtain other preferred implementations.

[0040] "Distillation" (distillation): A method of compressing the knowledge of a large network into a small network. First train a large network, use the appropriate temperature parameter T in the final softmax layer, and the probability obtained from the final training is called "soft target". Take this soft target and real label as the target to train a relatively small network, and also use the temperature parameter T determined in the...

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 belongs to the technical field of information processing, and particularly relates to a search method. At present, a relatively mature scheme generally has an NAS (Neural Architectural Searching) mode, an autonomous optimization platform AotuML provided by a Google company and the like, but still needs to consume a large amount of computing resources. The invention provides a constraint-based network search method. The method comprises the following steps: firstly, obtaining a constraint method; obtaining a search method based on a constraint method; and finally, obtaining a complete constrained deep network autonomous search method and a matched framework thereof. The precision as high as possible is realized under the specified network complexity; or the network complexitydesign as low as possible is realized above the specified precision; and the application of the constraint mechanism in autonomous search is realized.

Description

technical field [0001] The present application belongs to the technical field of information processing, and in particular relates to a search method. Background technique [0002] Deep Neural Networks (hereinafter referred to as DNN) is the basis of deep learning. DNN can be understood as a neural network with many hidden layers, also known as deep feedforward network (DFN), multi-layer perceptron (Multi-Layer perceptron, MLP). The traditional unoptimized deep neural network has a large number of parameters, and training and testing consume more computing resources and computing time, which is not conducive to actual research; it cannot achieve directional optimization, and can only rely on manual repeated attempts, which is error-prone and labor-intensive inefficiencies; [0003] The optimized deep neural network uses reinforcement learning or evolutionary learning to achieve network optimization. The optimized network itself needs more resources to realize when it is in...

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): G06F16/903G06N3/04G06N3/08
CPCG06F16/903G06N3/082G06N3/045
Inventor 张昱航叶可江须成忠
Owner SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI