Student competition

Three categories

• Minimization

• Nonlinear systems

• Least Squares problems

Objective of the competition:

1. Solve the maximum number of problems

2. Minimize the cost of the solution

Competition timetable:

• Competition begin at 9.00 on February 20, 2012

• 10 problems for each category are proposed to the competitors

• Student must write a matlab function which solve each problem

• Function must have signature like: function x=MySolution(x0,fun) where x(:,i) is the i-th iterate, x0 the starting point and fun the function to be minimize

• Competition end at 21.00 on March 2, 2012

• End of the competition is postponed at 24.00 on March 18, 2012

Competition ranking formation

• Each competitor gets a point for each problem his program is able to solve

• “undisclosed” problem (for each category) are use in rank formation

• The winner is the competitor which maximize points (maximize number of problems solved)

• In case of equal score with 2 or more competitor a computational cost index is used:

1. For each problem an additional point is attributed to competitor that minimizes the index: nf + n * ng + n^2 * nh where nf number of function evaluation, ng number of gradient evaluation, nh number of hessian evaluation.

2. In case of nonlinear system the index is: nF + n * nJ where nF number of map evaluation, nJ number of jacobian evaluation.

• In case equal score (very unlikely) additional and more difficult problems are considered until a winner is found

Computational limitation for each problem

• n is the number of unknowns

• A limit of 1000 iteration

• A limit of n*n*1000 function evaluations

• A limit of n*1000 gradient evaluations

• A limit of 1000 hessian evaluations

• Competition opened!

text of competition (UPDATED)

---