Lecture 12.2: gradient twisting approaches at their best: Newton's method
Completion requirements
The next step: changing the direction ... and there are very many to choose from: the Twisted Gradient Method, Zoutendijk's result. The "most obvious" one: Newton's method. Basic ideas for the (strongly) convex case, expected behavior. MATLAB implementation of Newton's method (basic version), first look at the behavior on some test cases. Take away: convergence is very good even far from the local minimum and indeed exceptionally good "in the tail".