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

Liquid state machine online learning method combining unsupervised learning and supervised learning

A liquid state machine, supervised learning technology, applied in neural learning methods, neural architectures, biological neural network models, etc., can solve the problems of slow parameter convergence, loss of dynamic targets, obstacles, etc., to achieve the effect of improving training speed

Pending Publication Date: 2022-04-22
SUN YAT SEN UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] 1. LMS-FORCE learning is a kind of gradient descent FORCE learning. Although the computational complexity is low and it is biologically reasonable, it has the problems of slow parameter convergence and low stability of the trained network.
[0006] 2. The randomly generated reserve pool is often not optimal. The reserve pool remains unchanged after being randomly generated. The learning method that only trains the output weight will be hindered by forgetting, and the learning will lose the previous dynamic goal after being disturbed. Causes a deviation from the target state immediately after the training

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
  • Liquid state machine online learning method combining unsupervised learning and supervised learning
  • Liquid state machine online learning method combining unsupervised learning and supervised learning
  • Liquid state machine online learning method combining unsupervised learning and supervised learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0055] In order to make the purpose of the present application, the technical solution and the advantages more clearly understood, the following combined with the accompanying drawings and embodiments, the present application will be further detailed in detail. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0056] The relevant technical terms involved in the present invention are first explained below:

[0057] Recurrent neural network: it is an artificial neural network model with signal feedback function, whose cyclic topology can maintain the self-continuous activation of the network, and save the historical input information in the internal state signal after nonlinear conversion, that is, it has dynamic short-term memory. Recurrent neural networks have been shown to approximate a variety of complex dynamic systems with arbitrary precision. Therefore, r...

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 discloses a liquid state machine online learning method combining unsupervised learning and supervised learning. The method comprises the following steps: constructing a spontaneous chaotic liquid state machine model with a feedback loop; constructing an STDP learning rule, and according to the STDP learning rule, carrying out adaptive optimization on a reserve pool synaptic weight of the liquid state machine model; and training the output weight of the liquid state machine according to LMS-FORCE learning based on a memory regression extension technology to obtain a trained liquid state machine. The method can improve the training speed and the stability of the training result, and can be widely applied to the technical field of artificial intelligence.

Description

Technical field [0001] The present invention relates to the field of artificial intelligence technology, in particular a liquid state machine online learning method that combines unsupervised and supervised learning. Background [0002] The traditional liquid state machine model consists of a randomly fixed reserve pool as a generic pretreatment core and a trainable output layer that extracts a high-dimensional active state that is stored in the reserve pool after a nonlinear transformation to map to the corresponding output. The core of the liquid state machine is to use the recursive structure of the reserve pool as a general filter, mapping the input sequence to a high-dimensional reserve pool state. Sparse and randomly interconnected reserve pools exhibit highly nonlinear dynamics, enabling inputs to be converted into rich, high-dimensional representations based on past contexts. In addition, the circular connection of the reserve pool leads to the emergence of short-term me...

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): G06N3/08G06N3/04
CPCG06N3/08G06N3/088G06N3/0418G06N3/049G06N3/044
Inventor 潘永平李艳松
Owner SUN YAT SEN UNIV