A Generalized Approach to the Method of Steepest Ascent

This paper derives a general steepest-ascent expression to show that any “direction” in the function space can be considered to be a steepest ascent direction by selecting a proper distance metric index and/or variable scaling. Several search procedures can be thus considered to be special cases of the method of steepest ascent. An illustrative investigation shows that, in a qualitative sense, most metric indices perform comparably.