Fig. 6
From: Accelerated ensemble optimization using momentum methods
The average NPV increase vs. the number of simulations. Each column corresponds to a value of a distinct method, while each row to a \(\beta\) value. The run with the best final NPV in each panel is plotted with a dashed line. Note that for Momentum and Nesterov, setting \(\beta =0\) yields the standard gradient descent method.
