Data quantification method and device based on neural network and computer readable storage medium

A technology of neural network and quantization method, which is applied in the field of data quantification based on neural network and computer-readable storage media, can solve problems affecting quantization performance and not applicable to actual work scenarios, and achieve the effect of reducing quantization errors and improving quantization effects
CN111008701APending Publication Date: 2020-04-14CANAAN BRIGHT SIGHT CO LTD
0 Cites 6 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CANAAN BRIGHT SIGHT CO LTD
Publication Date
2020-04-14

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention provides a data quantification method and device based on a neural network and a computer readable storage medium. The method comprises the steps of determining the first output activation of a target layer of the neural network according to any one preset frame in a to-be-detected frame sequence, and carrying out the quantification of the first output activation based on a current quantification parameter; iteratively updating the current quantization parameter according to the first output activation of the target layer and the target quantization bit width; and determining second output activation of the target layer according to the next frame of any preset frame in the to-be-tested frame sequence, and performing quantization operation on the second output activation based on the updated current quantization parameter. By means of the method, quantization errors can be reduced, and a better data quantization effect is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention belongs to the field of neural network calculation, and in particular relates to a data quantification method, device and computer-readable storage medium based on a neural network. Background technique

[0002] This section is intended to provide a background or context for implementations of the invention that are recited in the claims. The descriptions herein are not admitted to be prior art by inclusion in this section.

[0003] In recent years, with the rapid development of neural networks, the computing performance of neural networks has been continuously improved. However, due to the shortcomings of large amounts of calculations, neural networks limit their deployment and use on edge devices or low-power devices. Data quantization is the solution to the above problems. One of the problem methods. However, in the prior art, quantization parameters used for performing quantization on neural network data may not be suitable for actu...

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