Hidden layer neuron adaptive activation method and device and terminal equipment

A neuron adaptive and neuron technology, applied in the field of hidden layer neuron adaptive activation, can solve the problems of complex calculation process, cumbersome hardware connection, and increased implementation cost, and achieve simple calculation process, simple hardware connection, and reduced implementation cost effect

Active Publication Date: 2020-08-25
NAT UNIV OF DEFENSE TECH
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] At present, the realization of lateral inhibition between neurons in the hidden layer requires the introduction of an additional inhibitory layer neuron, which makes the calculation process more complicated;

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
  • Hidden layer neuron adaptive activation method and device and terminal equipment
  • Hidden layer neuron adaptive activation method and device and terminal equipment
  • Hidden layer neuron adaptive activation method and device and terminal equipment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0045] For this example, see figure 1 , shows a hidden layer neuron adaptive activation method, the method includes the following steps:

[0046] Step S100: Calculate the average input current according to the input current of each hidden layer neuron in the same layer.

[0047] The activation mode of hidden layer neurons in this embodiment is derived from the classic leaky integral ignition (LIF) neuron model, on this basis, the average input current is calculated according to the input current of each hidden layer neuron in the same layer in Represents the input current of the mth hidden layer neuron in the lth layer, M represents the total number of hidden layer neurons in the lth layer in the neural network, represents the average input current.

[0048] Step S200: Inject the average input current into each hidden layer neuron.

[0049] Exemplary, see figure 2 , injecting the average input current calculated and obtained according to the input current of each hidd...

Embodiment 2

[0063] In the training phase, the activation of hidden layer neurons is not only affected by other hidden layer neurons in parallel with it, but also by itself in the time dimension. For this example, see image 3 , shows the interaction of the hidden layer neurons themselves in the time dimension in the hidden layer neuron adaptive activation method.

[0064] It can be understood that after the mth hidden layer neuron in the l layer enters the activation state, compare the current moment activation state of the discharged hidden layer neuron with the preset activation state threshold, when the current moment activation of the discharged hidden layer neuron When the state is less than or equal to the preset activation state threshold, update the current moment activation state of the hidden layer neuron of the discharge; when the current moment activation state of the discharge hidden layer neuron is greater than the preset activation state threshold, the discharge hidden laye...

Embodiment 3

[0078] For this example, see Figure 4 , shows a schematic structural diagram of the hidden layer neuron adaptive activation device 1 , which includes an initial module 100 , a cancellation module 200 , a discharge module 300 , a calculation module 400 and an activation module 500 .

[0079] The initial module 100 is used to calculate the average input current according to the input current of each hidden layer neuron in the same layer; the cancellation module 200 is used to inject the average input current into each hidden layer neuron; the discharge module 300, It is used to control the corresponding hidden layer neurons whose input current is greater than the average input current to discharge sequentially according to the preset discharge sequence rule; the calculation module 400 is used to determine the respective Activation values ​​of the firing hidden layer neurons; an activation module 500 , configured to use the activation values ​​to correspondingly activate the res...

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 embodiment of the invention discloses a hidden layer neuron adaptive activation method and device and terminal equipment applied to a memristor pulse neural network, each synapse in the memristorpulse neural network only comprises one memristor, and the method comprises the following steps: calculating an average input current according to the input current of each hidden layer neuron in thesame layer; injecting the average input current into each hidden layer neuron; controlling the hidden layer neurons of which the corresponding input current is greater than the average input current to discharge in sequence according to a preset discharge sequence rule; determining an activation value of each discharged hidden layer neuron according to the discharge sequence of each discharged hidden layer neuron; and correspondingly activating each discharged hidden layer neuron by utilizing the activation value. Lateral suppression between hidden layer neurons is realized; and each synapse can only comprise one memristor, so that the construction of a complex hardware connection network is avoided.

Description

technical field [0001] The present invention relates to the field of artificial intelligence, in particular to a hidden layer neuron adaptive activation method, device and terminal equipment. Background technique [0002] In the era of big data, artificial intelligence (AI) technology is developing rapidly. The memristor spiking neural network based on memristor (Memristor) as the physical basis and spiking neural network (SNNs, Spking Neural Networks) as the algorithm provides a high-efficiency brain-like computing solution. Memristor is a two-port non-volatile memory device whose resistance state can be changed by the voltage applied across it or the current flowing through it, with high integration density and extremely low operating power consumption; pulsed neural network is a Novel biologically-inspired networks that transmit information between neurons via spiking signals with spatiotemporal information have great potential for network efficacy. Since it has been pr...

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): G06N3/04
CPCG06N3/063G06N3/045Y02D10/00
Inventor 李楠李清江刘森李纪伟徐晖刁节涛陈长林宋兵王义楠刘海军于红旗李智炜王伟王玺步凯
Owner NAT UNIV OF DEFENSE TECH
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