Sign-direction conjugate gradient algorithm

This paper presents a new version of conjugate gradient (CG) algorithm, which depends mainly on the sign of direction of search, the algorithm is called SDCGA. In addition, the algorithm is simple so that the weight vector update depends on current input signal vector and it also avoids the internal loops which are essential to the conventional CG algorithm for the search direction update. It is also noticed that the input signal autocorrelation is not involved in any calculate updates. The algorithm is tested against the least mean square (LMS) and it proves it is faster and more accurate.