DocumentCode :
1477297
Title :
Simple and Effective Extrapolation Technique for Neural-Based Microwave Modeling
Author :
Zhang, Lei ; Zhang, Qi-Jun
Author_Institution :
RF Div., Freescale Semicond., Inc., Tempe, AZ, USA
Volume :
20
Issue :
6
fYear :
2010
fDate :
6/1/2010 12:00:00 AM
Firstpage :
301
Lastpage :
303
Abstract :
This letter presents a simple and effective technique allowing neural network based microwave models to be used outside their training range. Our technique detects necessity of extrapolation inside each hidden neuron and performs extrapolation using a modified neuron activation function. Compared to existing techniques, the proposed technique is easy to implement and the extrapolation result is equally good or better. Examples of extrapolating neural models of a high electron mobility transistor and a power amplifier are presented.
Keywords :
electronic engineering computing; extrapolation; high electron mobility transistors; microwave devices; neural nets; power amplifiers; extrapolating neural models; extrapolation technique; high electron mobility transistor; neural network based microwave models; neuron activation function; power amplifier; training range; Extrapolation; microwave modeling; neural networks; neuron;
fLanguage :
English
Journal_Title :
Microwave and Wireless Components Letters, IEEE
Publisher :
ieee
ISSN :
1531-1309
Type :
jour
DOI :
10.1109/LMWC.2010.2047450
Filename :
5453037
Link To Document :
بازگشت