Event-based maximum pooling method, chip and electronic product

A technology of maximum pooling and electronic products, applied in physical implementation, neural architecture, biological neural network models, etc., can solve the problems of undiscovered maximum pooling schemes, openness, etc., to achieve convenient reset, prevent overflow, and flexible accuracy tune effect

Active Publication Date: 2021-11-19
HENGDU SYNSENSE TECH CO LTD +1
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the inventor has not found any event-based maximum pooling scheme disclosed

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
  • Event-based maximum pooling method, chip and electronic product
  • Event-based maximum pooling method, chip and electronic product
  • Event-based maximum pooling method, chip and electronic product

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0040] Although the numbers of some steps are given in the embodiments, and these numbers have a numerical order of magnitude, this does not necessarily mean the absolute order of execution of the steps. The above numbering of the present invention does not imply an order of execution. Based on the concept of the present invention, the concept of the present invention will be described in detail below with some specific embodiments, which are only a part of the detailed description of the concept of the invention, and the claims will be intended to be based on these specific embodiments. Outlines the idea of ​​the invention.

[0041] refer to figure 2 , which shows the mapping relationship between events and counts. In an event-driven system, events are generated based on changes in event signals sensed by sensors, such as a common event camera—Dynamic Vision Sensor (DVS). If the signal received by the sensor does not change, no event is generated, so no event is passed to...

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 designs an event-based maximum pooling method, a chip and an electronic product. The method comprises the following steps of determining a pooling target according to an input event, and the pooling target being an event set; loading a plurality of count values corresponding to the pooling target; increasing a count value corresponding to the input event to obtain an updated count value corresponding to the input event; and judging whether the count value corresponding to the updated input event is the maximum value in the plurality of count values corresponding to the pooling target, if so, sending the input event, and if not, discarding the input event. In order to solve the technical problem that a maximum pooling scheme for the event-driven system does not exist at present, the event-based maximum pooling scheme is proposed for the first time, and the technical effects of flexible and adjustable precision, overflow prevention, convenience in resetting and the like are also achieved.

Description

technical field [0001] The invention relates to an event-based maximum pooling method, a chip and an electronic product, in particular to a method for realizing maximum pooling in a pulse neural network of a brain-like chip, a chip and an electronic product. Background technique [0002] Operations such as convolution and pooling often appear in convolutional neural networks, and maximum pooling is the most common form of pooling. It is extremely easy to compare a set of values ​​and select the maximum value in the traditional von Neumann computer architecture, but it is much more difficult in an event-driven system. [0003] refer to figure 1 , which is a schematic diagram of the artificial neural network (Artificial Neural Network, ANN) maximum pooling process in the prior art. For a 4×4 feature map matrix, it is scanned by a 2×2 filter (also called a pooling kernel) with a step size of 2, and the final output result is a 2×2 feature map. such as reference figure 1 In ...

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): G06N3/04G06N3/063
CPCG06N3/049G06N3/063
Inventor 乔宁邢雁南柯政西克·萨迪克·尤艾尔阿明白鑫周凯任宇迪张林范健均
Owner HENGDU SYNSENSE 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