Steepest-descent optimization method

Question

I was wondering if the steepest descent method can find a global min/max or only local min/max?

Thanks

Answer 1

There is no guarantee that a steepest descent will always climb down to a global minimum.

This depends on the starting point of the search.

from http://www.mathworks.com

Imagine two adjacent valleys. You have to take the right direction to end up in the deeper one.

Ask Google for Rosenbrock function.