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

Stability maintaining method in echo state network size compression

An echo state network, stability technology, applied in the field of machine learning, can solve problems such as changing network stability

Pending Publication Date: 2022-05-10
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The present invention provides a method for maintaining stability in the compression of the echo state network, and its purpose is to solve the problem that the network stability may be changed when the echo state network is compressed in network size

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
  • Stability maintaining method in echo state network size compression
  • Stability maintaining method in echo state network size compression
  • Stability maintaining method in echo state network size compression

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0008] In order to better illustrate the technical implementation process of the present invention, each technical part will be described more clearly below. The examples described in the present invention are only a part, not all applicable examples. Based on the examples in the present invention, all other examples obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0009] have n in input, n out The ESN of n outputs and n internal variables can be expressed by the following nonlinear difference equation in discrete-time k-state space form:

[0010]

[0011] in, is the state vector of n internal units at time k, and are the input and output vectors at time k, is the input weight matrix for the connection between input units and internal units, is the internal weight matrix of internal units connected from time k to k+1, is the output weight matrix of connections between in...

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 the field of machine learning, in particular to a stability maintaining method in echo state network size compression. The invention aims to provide a method for keeping stability in order to solve the problem that network stability may be changed when network size compression is carried out on an echo state network. The method comprises the following steps: firstly, decomposing an activation function of an original echo state network into two parts at a balance point through Taylor expansion, namely a nonlinear part with a Jacobian matrix being zero and a linear part of a Jacobian matrix inheriting the original activation function; and performing operation such as function approximation on the nonlinear part without changing a Jacobian matrix of the whole function at a balance point. Therefore, the size of the network is compressed, the forward propagation speed of the network is accelerated, and the same stability as that of the original network is kept.

Description

technical field [0001] The invention relates to the field of machine learning, in particular to a method for maintaining stability in echo state network size compression. [0002] technical background [0003] The Echo State Network (ESN) belongs to the Recurrent Neural Network (RNN), which is a deep network that specializes in processing temporal data. It is an important memory-based computing neural network. ESNs are known for their simple structure and large internal recursive topology called reservoirs. In addition, unlike ordinary RNN whose weights can be trained, the linear regression method can only train the output weights of the echo state network, which avoids the long-term dependence problem in ordinary RNN training. Echo state networks have been widely used in many time series and nonlinear dynamical system modeling applications due to their simple structure and training process based on linear regression. [0004] After training, a neural network will be used ...

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): G06N3/04G06N3/08
CPCG06N3/082G06N3/048G06N3/044
Inventor 王海龙行毅郭宪章
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA
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