An optimum random-search algorithm is considered. The convergence conditions to the greatest increase (local properties) and convergence to the point of extremum (integral properties) of a function by optimizing in the presence of noise, are found. The results are used for finding a global...