Maximize multivariable function with infinite maxima

Join the discussion
Registration is free. Start your own thread to ask a follow-up.
3 replies · 2K views
Physics news on Phys.org
Well, if you know that the set of solutions form a circle, then you can transform the problem to one dimension by changing to a polar coordinate system, no?
 
I want to be able to solve in Cartesian coordinates, I think, since this is the very simplified form of a function which will contain many more terms.
 
In general, you can try the second partial derivative test.

Let [itex]\vec H_k(f(\vec x))[/itex] be the Hessian matrix of the function [itex]f(\vec x)[/itex] (evaluated at [itex]\vec x[/itex]) of the [itex]k[/itex] first variables, where [itex]k = 1, 2, 3, ... , n[/itex].

If you're function is [itex]f(\vec x)[/itex] then the critical point [itex]\vec p[/itex], i.e. [itex]\nabla f(\vec p) = \vec 0[/itex], is a local minimum if [itex]\forall k : |\vec H_k(f(\vec p))| > 0[/itex] and a local maximum if [itex]\forall k : (-1)^k |\vec H_k(f(\vec p))| > 0[/itex]. For all other cases, [itex]\vec p[/itex] is a saddle point unless [itex]|\vec H_n(f(\vec p))| = 0[/itex], for which the test is inconclusive.