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

Data processing system and method

A data processing and processor technology, applied in the field of neural networks to ensure the effect of output accuracy

Active Publication Date: 2019-06-21
NANJING HORIZON ROBOTICS TECH CO LTD
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] However, in order to ensure the output accuracy of the embedded device, it is possible to provide input data (eg, an image with high resolution) with a data volume larger than the capacity of the SRAM

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 system and method
  • Data processing system and method
  • Data processing system and method

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0093] refer to figure 2 In the illustrative example of , it is assumed that data A occupies 40% of the memory cell capacity, data B occupies 15% of the memory cell capacity, data C occupies 40% of the memory cell capacity, data D occupies 70% of the memory cell capacity, and data E 20% of the memory cell capacity. In this example, operation 2013 may be identified as a first operation whose amount of input data C and output data D exceeds the capacity of the memory cell (eg, 40%+70% > 100%).

[0094] In some embodiments, the operation 2013 identified for which the amount of input data C and output data D exceeds the capacity of the memory cell may be grouped with at least one second operation selected from a series of operations. The at least one second operation may be as Figure 5Operation 5012 in the grouping scheme shown. In this grouping scheme, grouping operations 5012-5013 may have total input data B and total output data D. The remaining storage space 5050 of the ...

example 2

[0102] In another illustrative example, refer to figure 2 , assuming that data A occupies 40% of the capacity of the memory cell, data B occupies 40% of the capacity of the memory cell, data C occupies 80% of the capacity of the memory cell, data D occupies 80% of the capacity of the memory cell, and data E occupies 10% of the capacity of the memory cell %. In this example, operation 2012 may be identified as the first operation from a series of operations 2011-2014 for which the amount of input data and output data exceeds the capacity of the memory cell (eg, 40%+80% > 100%).

[0103] When an operation 2012 is identified, a second operation may first be selected from the series of operations, wherein the second operation will be grouped with the identified operation 2012 . In this example, there are two possible grouping schemes, and there is only one second operation to be grouped with operation 2012 . For example, operation 2011 may be grouped with operation 2012 , or op...

example 3

[0108] In another illustrative example, refer to figure 2, assuming that data A occupies 10% of the capacity of the memory cell, data B occupies 40% of the capacity of the memory cell, data C occupies 80% of the capacity of the memory cell, data D occupies 80% of the capacity of the memory cell, and data E occupies 10% of the capacity of the memory cell %. In this example, operation 2012 can be identified as the first operation from a series of operations 2011-2014, the amount of input data and output data of operation 2012 exceeds the capacity of the memory cell (eg, 40%+80%>100%).

[0109] When an operation 2012 is identified, a second operation may first be selected from the series of operations, wherein said second operation is to be grouped with the identified operation 2012 . In this example, there are two possible grouping schemes, and there is only one second operation to be grouped with operation 2012 . For example, operation 2011 may be grouped with operation 2012...

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

A systems and method of data processing are provided. The method includes following steps of receiving input data to be processed by a series of operations; identifying a first operation from the series of operations; selecting at least one second operation from the series of operations to group with the first operation, which is based at least in part on the amount of input data and output data of the group operation and the capacity of the memory cell; and processing a portion of the input data of the set of operations. By ensuring that input data and output data of any data operation are stored in the memory unit, the efficiency of the series of data operations can be improved.

Description

technical field [0001] The present application relates to neural network technology, especially a data processing system and method. Background technique [0002] Deep computing frameworks such as convolutional neural networks (CNNs) have been used in many application domains, including pattern recognition, signal processing, time series analysis, etc. With the development of convolutional neural network algorithms, CNN can be used in more and more application scenarios. Some application scenarios of CNNs may be sensitive to power consumption and cost. For example, the capacity of Static Random Access Memory (SRAM) of an embedded device is limited to reduce the cost of the embedded device. [0003] However, in order to ensure the output accuracy of the embedded device, the amount of input data (for example, an image with high resolution) may be provided larger than the capacity of the SRAM. It would be beneficial to have methods and systems that implement operations by sp...

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): G06F9/50G06N3/063
CPCG06F9/5016G06N3/063G06N3/048G06N3/045G06F9/5061G06F9/3004G06F9/30007G06F9/30076G06F9/345G06F9/4881
Inventor 王振江李建军陈亮凌坤李德林孙晨
Owner NANJING HORIZON ROBOTICS TECH CO LTD