Gradient Ascent is Gradient Descent but with the opposite objective, rather than minimizing you maximize.

This is used when trying to maximize reward