The problem is considered of maximizing a function in a convex region. To solve this problem a new method is developed, to be called "method of feasible directions". It is a method of steep ascent.