KNITRO KNITRO is an optimization software library for finding solutions to both regular and continuous (constrained and unconstrained) optimization models, as well as various optimization models with integer or binary variables (i.e., mixed integer programs). KNITRO is primarily designed to find large-scale local optimal solutions to continuous nonlinear problems.

The software is designed for large problems with dimensions in the hundreds of thousands. It is effective for solving regular linear, quadratic, or nonlinear optimization problems, both convex and nonconvex. It is also effective for nonlinear regression, problems with complementarity constraints (MPCC or MPEC), and mixed integer programming (MIP), mixed convex and nonlinear integer problems (MINLP). It is highly regarded for its robustness and efficiency.

Count on three state-of-the-art algorithms/solvers for problem solving. Each algorithm addresses the set of nonlinear optimization problems, and each is designed to achieve maximum efficiency at large scale.

The Direct Interior-Point algorithm applies barrier techniques and directly factors the KKT matrix of the nonlinear system. It performs best on ill-conditioned problems.

The interior-point CG algorithm applies barrier techniques, using the conjugate gradient method, to solve KKT subproblems. It provides an alternative to the direct interior-point algorithm when KKT factorization is impractical or inefficient.

Active Set combines classical active set principles with a linear programming subproblem to rapidly discover the set of binding constraints. Its behavior is significantly different from that of interior point algorithms and converges accurately to the active set of high-precision sensitivity information.

Having three algorithms also means that KNITRO's interior point solutions can be "cross-referenced" to the Active Set algorithm for the final calculation of high-precision solutions and sensitivities. Some products attempt a heuristic-based cross-reference solution, but only a rigorous Active Set method can determine the correct active set with true confidence.

Every nonlinear optimization problem is unique, and it can be quite difficult to predict the performance of any algorithm on a given problem. Choosing from three KNITRO algorithms greatly increases the chances of solving your problem efficiently. KNITRO provides a wide range of user options and offers interfaces for C, C++, Fortran, Java, Python, AMPL, AIMMS, GAMS, MPL, MATLAB, Microsoft Excel, and LabVIEW. Active and ongoing development and support ensure that the software will remain a leader in nonlinear optimization.