Oh, that I would approach analytically maybe? In 1D on a plateau you would have the second derivative to be zero, for a strictly convex function, you would not have that. You can do the same in the multivariate case looking at Hessians.
Oh, that I would approach analytically maybe? In 1D on a plateau you would have the second derivative to be zero, for a strictly convex function, you would not have that. You can do the same in the multivariate case looking at Hessians.