A neural network model for invariant pattern recognition


It is shown that a condition for linear stability of the LMS<lb>algorithm given by Leung and Haykin’ is not satisfied by any matrix. In the above paper,’ Leung and Haykin state that the LMS al-<lb>gorithm will converge linearly if the Euclidean norm of I -<lb>puM(n)u&(n)<lb>is less than one. This statement is true; unfortunately<lb>this condition is not… (More)
DOI: 10.1109/78.139270

3 Figures and Tables


  • Presentations referencing similar topics