Linear and nonlinear associative memories for parameter estimation (Q1183172)

This paper proposes the use of associative memories for obtaining preliminary parameter estimates for nonlinear systems. When a training set \((r_ i,s_ i)\) is given an associative memory matrix \(\widehat M\) is constructed which optimally, in the least squares sense, associates each system output vector \(s_ i\) with its corresponding parameter vector \(r_ i\). Given any observed system output vector \(s^*\), an estimate \(\widehat r\) for the system parameters is obtained by setting \(\widehat r= \widehat M s^*\). This paper also gives numerical experiments to indicate the effectiveness of this approach.