Lecture 4.2: the gradient method for quadratic functions, practice
Aggregazione dei criteri
MATLAB implementation of the gradient method for quadratic functions. Running the code on different test instances, observing the properties: directions hortogonal, convergence rate dependent on shape of ellipsoids = eigenvalues. Differences between a pseudo-code and real code, special cases (unbounded problems, some "easy to spot" and some not), streamlining. Take away: the algorithm seems to work, but it can do so in very different ways and there is the need to understand why.