Frank-Wolfe Algorithm - Final Iteration Previous Next
Third trial solution:
Evaluating the partial derivatives at (1.695, 1.914):
so the approximating objective function is .
Maximizing , or equivalently (after dividing by 4.17)
 , results in every solution on the line between (2, 1) and (0, 7) being optimal for this linear programming problem. Regardless of which solution is used as the other endpoint of the line from (1.695, 1.914), we already know from the preceding iteration that  is maximized along this line at (1.695, 1.914).
Since the trial solution did not move, this verifies that the optimal solution for our convex programming problem is
This concludes the demonstration. See the OR Tutor menu (to the left) for other demonstrations or close the browser window to exit OR Tutor.