Improvements of Analog Neural Networks Based on Kalman Filter
In the paper, original improvements of recurrent analog neuralnetworks, which are based on Kalman filter, are presented. Theseimprovements eliminate some disadvantages of the classical Kalmanneural network and enable a real time processing of quickly changingsignals, which appear in adaptive antenna...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Spolecnost pro radioelektronicke inzenyrstvi
2002-04-01
|
| Series: | Radioengineering |
| Online Access: | http://www.radioeng.cz/fulltexts/2002/02_01_06_13.pdf |
| Summary: | In the paper, original improvements of recurrent analog neuralnetworks, which are based on Kalman filter, are presented. Theseimprovements eliminate some disadvantages of the classical Kalmanneural network and enable a real time processing of quickly changingsignals, which appear in adaptive antennas and similar applications.This goal is reached using such circuit elements, which increase theconvergence rate of the network and decrease the dependence ofconvergence rate on the ratio of eigenvalues of the correlation matrixof input signals. |
|---|---|
| ISSN: | 1210-2512 |