Mixed gradient based fast learning algorithm for MLP with its applications to MQAM digital mobile radio receptions

A new fast adaptive learning algorithm called a mixed gradient based fast learning algorithm with variable learning gain and selective weight updates is proposed which uses the steepest descent method combined with the conjugate gradient method. The line search unconstrained optimization method is adopted to adaptively adjust the learning gain eta . The derivation of the fast learning algorithm is given. Computer simulation results are included.<>