Results can be found at the bottom of the page (shows points, times and the number of instances not solved): many got the full points but since there needs to be one winner... it is jokinen who had the fastest algorithms (nobody got all instances within 0.01 but they were really close to that)
Preliminary deadline: Monday Feb 6th, 6pm
Return one zip file that contains two files: studenta.m and studentb.m where "student" is your last name.
The functions take two inputs: the minimized function and the initial solution as a column vector.
Your assignment is to design two algorithms with MATLAB for unconstrained minimization: a) one-dimensional local search and b) multidimensional local search. You cannot use MATLAB's own optimization algorithms, but external mex/java/etc-files of your own can be used. The performance of the algorithms is tested by measuring the run time against MATLAB's own and other students' algorithms. For both algorithms, the tests are made with three different functions with ten different initial values (and the average time is counted from these). Who wins the tournament and who will program the best algorithm?
The tolerance for the optimization variable is 10-2 in maximum norm. If the algorithm returns a point that is close enough and satisfies this condition, then it is accepted. If the algorithm returns a point further away from the optimum, it is penalized in the run time. Therefore, the evaluation of the algorithm is multicriteria with respect to time and the distance from the optimum.
Write functions that take the minimized function and the initial value as the input so that they are of form function(f,x0) and it returns the solution x, where all the values are column vectors. Name the functions according to your last name and in the end a or b, e.g., johnsona.m and johnsonb.m. The function f to be minimized takes as input one column vector. For example, f=@(x) x.^2; or f=inline('x.^2'); johnsona(f,2). Return the MATLAB files electronically to MyCourses as a one zipped file.
Please add a one-line comment (using %) at the very beginning of both matlab-files that is a short description of your algorithm.