In this seminar, we wish to introduce an Ampl-based optimization problem analyzer and discuss its potential when used either as a stand-alone tool or when hooked to a server such as NEOS. The capabilities of this analyzer include classification of the problem at hand, analysis of the objective and constraint function and their variables, the providing of upper and lower bounds on the values of these functions on the feasible set and assessment of convexity. Two antagonistic approaches are considered for the latter point; namely a convexity disprover and a convexity prover. Our approach is compared to existing work.

Eventually, we expect this analyzer to create a link between a list of available solvers and the class of problems they treat and provide aid in the choice of an appropriate solver for a given problem.