Filling parameter optimization method and device, and computer readable storage medium

A parameter and numerical technology, applied in the field of optimization of filling parameters, can solve the problems of multi-functionality of filling methods, the inability to realize various fillings, and the inability to be set, and achieves the effect of improving operation speed, reducing input scale, and reducing memory usage.

Inactive Publication Date: 2020-05-08
CAMBRICON TECH CO LTD
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] However, filling (zero padding) cannot be set under the TensorFlow framework, so there are many functional defects in the filling method, and various filling situations cannot be realized.
In the Pytorch framework, it is allowed to set the padding parameter to any non-negative integer value. There is no functional defect of the TensorFlow framework, but because the same padding or zero padding is performed at both ends of the corresponding dimension, there is a phenomenon of redundancy.

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
  • Filling parameter optimization method and device, and computer readable storage medium
  • Filling parameter optimization method and device, and computer readable storage medium
  • Filling parameter optimization method and device, and computer readable storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0096] Apply the above example of redundant phenomenon caused by the filling method of the prior art to an embodiment of the present disclosure to see whether there is redundant phenomenon.

[0097] Assume that in Example 1, each value of the height (H) dimension is equal to each value of the width (W) dimension, and the corresponding parameters in the input data are:

[0098] Step size (stride)=3, convolution kernel expansion parameter (dilation)=1, input size (input_size)=7, convolution kernel size (filter_size)=4, filling parameter (padding)=2,

[0099] The input data (input) is:

[0100] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

[0101] Table 6

[0102] Then, the expanded size of the convolution kernel (dilation_filter_size)=dilation×(filter_size-1)+1=1×(4-1)+1=4

[0103] The expanded convolution kernel:

[0104] 1 1 1 1 ...

Embodiment 2

[0123] Assume that in Example 2, each value of the height (H) dimension is equal to each value of the width (W) dimension, and the corresponding parameters in the input data are:

[0124]Step size (stride) = 2, convolution kernel expansion parameter (dilation) = 1, input size (input_size) = 7, convolution kernel size (filter_size) = 4, filling parameter (padding) = 2,

[0125] The input data (input) is:

[0126] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

[0127] Table 12

[0128] Then, the expanded size of the convolution kernel (dilation_filter_size)=dilation×(filter_size-1)+1=1×(4-1)+1=4

[0129] The expanded convolution kernel:

[0130] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

[0131] Table 13

[0132] Filled input size (input_with_padding_size) = input_size+2×padding=7+2×2=11

[0133] Filled input data (in...

Embodiment 3

[0150] Assume that in embodiment 3, each value of the height (H) dimension is equal to each value of the width (W) dimension, and the corresponding parameters in the input data are:

[0151] Step size (stride) = 4, convolution kernel expansion parameter (dilation) = 1, input size (input_size) = 7, convolution kernel size (filter_size) = 4, filling parameter (padding) = 2,

[0152] The input data (input) is:

[0153] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

[0154] Table 18

[0155] Then, the expanded size of the convolution kernel (dilation_filter_size)=dilation×(filter_size-1)+1=1×(4-1)+1=4

[0156] The expanded convolution kernel:

[0157] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

[0158] Table 19

[0159] Filled input size (input_with_padding_size) = input_size+2×padding=7+2×2=11

[0160] Filled input data...

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 filling parameter optimization method and device, and a computer readable storage medium. The filling parameter optimization method can be included in a combined processingdevice. The combined processing device can further comprise a universal interconnection interface and other processing devices. The filling parameter optimization device interacts with the other processing devices to jointly complete calculation operation specified by a user. The combined processing device can further comprise a storage device; the storage device is connected with the filling parameter optimization device and the other processing devices and is used for the data service of the filling parameter optimization device and the other processing devices. With the filling parameter optimization method and device, and the computer readable storage medium of the invention adopted, the generation of a redundancy phenomenon is eliminated, an input scale is reduced, memory occupation is reduced, and operation speed is improved.

Description

technical field [0001] This disclosure relates generally to the field of computers. More specifically, it relates to a method, device, and computer-readable storage medium for optimizing filling parameters. Background technique [0002] Tensorflow is a symbolic mathematics system based on dataflow programming, which is widely used in the programming of various machine learning algorithms. Pytorch is the python version of torch, a neural network framework open sourced by Facebook. Different from Tensorflow's static calculation graph, Pytorch's calculation graph is dynamic, and the calculation graph can be changed in real time according to the calculation needs. In the Tensorflow framework and the Pytorch framework currently used in the field of artificial intelligence, the addition method of padding (zero padding) is often used in convolution operations. [0003] However, padding (zero padding) cannot be set under the TensorFlow framework, so there are many functional defe...

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): G06N3/063G06N3/04G06N3/08
CPCG06N3/063G06N3/08G06N3/045
Inventor 不公告发明人
Owner CAMBRICON 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