Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Neuron circuit circuit, chip, system, method and storage medium

A technology of neurons and neuron arrays, applied in the computer field, can solve the problems that the neural network architecture cannot be reconfigured and the application of deep learning chips is limited, and achieve the effect of expanding applications

Active Publication Date: 2019-03-01
深圳市中科元物芯科技有限公司
View PDF10 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The object of the present invention is to provide a neuron circuit, chip, system and its method, storage medium, aiming to solve the problems existing in the prior art, neural network architecture The problem that the application of deep learning chips is limited due to the inability to reconfigure

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
  • Neuron circuit circuit, chip, system, method and storage medium
  • Neuron circuit circuit, chip, system, method and storage medium
  • Neuron circuit circuit, chip, system, method and storage medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0047] figure 1 The structure of the neuron circuit provided by Embodiment 1 of the present invention is shown, specifically, it relates to a digital neuron circuit used to form a deep learning neural network, and the deep learning neural network can perform various required functions on the input data. The sequential processing of the neural network layer, and the neuron circuit is used to perform the data processing required on the corresponding node of the neural network layer. For ease of description, only parts related to the embodiments of the present invention are shown, including:

[0048]The calculation module 101 can adjust the calculation infrastructure to perform data processing of different neural network layer nodes. In this embodiment, the computing module 101 can be used to perform single processing such as corresponding multiplication, addition, and activation using an activation function, or to perform flexible combinations of different processing. The comp...

Embodiment 2

[0053] This embodiment further provides the following content on the basis of the neuron circuit in other embodiments:

[0054] Such as figure 2 As shown, the neuron circuit also includes:

[0055] The parameter storage module 201 is configured to store parameters required for data processing of neural network layer nodes. In this embodiment, the parameters may be neural network parameters obtained through training.

[0056] The address generation module 202 is configured to be controlled by the control module 103 to search for parameters corresponding to the data for neural network layer node data processing, and the searched parameters will be input to the calculation module 101 to participate in corresponding data processing.

[0057] In this embodiment, since the data processing of certain types of neural network layers does not need to call parameters, for example: pooling network layer, activation function network layer, etc., the basic configuration of neuron circuit...

Embodiment 3

[0059] This embodiment further provides the following content on the basis of the neuron circuit in other embodiments:

[0060] Such as image 3 As shown, the neuron circuit also includes:

[0061] The temporary storage module 301 is used for storing the intermediate data processed by the nodes of the neural network layer.

[0062] In this embodiment, since some types of neural networks, such as convolutional networks and regional networks, do not need to save the intermediate data processed by the neuron circuit for subsequent processing, the basic configuration of the neuron circuit does not require the above-mentioned temporary storage module 301, while other data processing such as reinforcement learning network and recurrent network need to use the intermediate data processed by neuron circuit, then the above temporary storage module 301 needs to be configured in the neuron circuit, which can also enhance the wide applicability of the neuron circuit.

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 is applicable to the computer technical field and provides a neuron circuit, a chip, a system and a method thereof, and a storage medium. The neuron circuit comprises the following structures: a calculation module; a configuration information storage module, configured to store configuration information of a neuron processing mode; and a control module, configured to control the computing module to adjust to the corresponding computing infrastructure and execute the corresponding neural network layer node data processing according to the processing mode configuration information.In this way, it can meet the fast iteration of complex and diverse neural network computing needs, can be widely used in computing resources are limited, the need for a certain neural network architecture reconfigurable areas, expand the depth of learning chip applications.

Description

technical field [0001] The invention belongs to the technical field of computers, and in particular relates to a neuron circuit, a chip, a system and a method thereof, and a storage medium. Background technique [0002] In recent years, with the wide application of artificial neural network-based deep learning technology in computer vision, natural language processing, intelligent system decision-making and other fields, the artificial intelligence chip technology that accelerates neural network calculations has attracted the attention of academia and industry. Pay attention to. [0003] Most of the existing ASIC (Application Specific Integrated Circuit, ASIC) chips customized for neural network computing are still based on pre-specified network structures and algorithms, excessively pursuing power consumption and speed performance, resulting in fixed hardware structures and no neural network capabilities. The reconfigurability of the architecture makes it impossible to dep...

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
IPC IPC(8): G06N3/063G06N3/04G06N3/08
CPCG06N3/063G06N3/08G06N3/045
Inventor 王峥梁明兰林跃金李善辽赵玮
Owner 深圳市中科元物芯科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com