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

A Compatible Neural Network Accelerator and Data Processing Method

A neural network and accelerator technology, applied in the field of computing, can solve problems such as limiting the universality of neural network accelerators

Active Publication Date: 2020-11-10
INST OF COMPUTING TECH CHINESE ACAD OF SCI
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Due to the differences in the calculation process of the two networks, it is usually necessary to use a dedicated neural network accelerator for data processing, which limits the universality of the neural network accelerator

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
  • A Compatible Neural Network Accelerator and Data Processing Method
  • A Compatible Neural Network Accelerator and Data Processing Method
  • A Compatible Neural Network Accelerator and Data Processing Method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the compatible neural network accelerator and data processing method provided in the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0024] The main time occupation of the CNN calculation process is the convolution process, which is mainly to multiply and accumulate the input neurons or pixels with the weights of the corresponding convolution kernels; the main time occupation of the LSTM network calculation process is the gate value Vector multiplication and accumulation operation and iterative operation process of each layer. This process is mainly to perform multiplication and accumulation of gate value vector and weight value, transfer the result as the input of the next layer of network, and perform the same operation.

[0025] In order to meet the needs of the above two network structures and ...

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 compatible neural network accelerator, comprising a storage unit for storing and outputting neuron data, weight data and control instructions; a matrix operation unit for receiving data from the storage unit according to the control instructions and Perform matrix operations on the received data and output operation results; a mode operation unit includes a plurality of functional modules, and the function modules can be used to obtain from the matrix operation unit and / or the activation unit and / or the storage unit or receiving data, performing a specific operation corresponding to the network according to the control instruction, and outputting an operation result; an activation unit, configured to receive data from the mode operation unit and / or the storage unit, and The received data performs an activation operation and outputs an activation result.

Description

technical field [0001] The invention relates to the computing field, in particular to an LSTM and CNN-oriented compatible neural network accelerator and a data processing method. Background technique [0002] Neural network is one of the perception models with a high level of development in the field of artificial intelligence. Once it appeared, it became a research hotspot in academia and industry. With the deepening of research, different types of neural networks have been proposed one after another. For example, convolution Neural network (CNN, Convolutional Neural Network) and long short-term memory network (LSTM, LongShort-Term Memory). [0003] CNN is a feed-forward neural network that includes a convolutional layer, often used in speech recognition, face recognition, motion analysis, natural language processing, etc.; LSTM network is a time-recursive cyclic neural network, often used in Learn about language translation, robot control, image analysis, and more. Due t...

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): G06N3/04G06N3/063
CPCG06N3/063G06N3/045
Inventor 韩银和闵丰许浩博王颖
Owner INST OF COMPUTING TECH CHINESE ACAD OF SCI