KNITRO 9.1 also brings new features in its MATLAB version. And thanks to our continuous efforts on improving performance, KNITRO 9.1 is 24% faster than KNITRO 9.0 on our benchmark models. Expect significant speedups for NLPs, MINLPs, and models with range constraints.
New KNITRO 9.1 detailed features:
New SQP algorithm designed for Derivative-Free Optimization:
- Tends to converge in fewest function evaluations,
- Good for warm-start,
- Can make use of the exact Hessian.
New features in MATLAB:
- Specialized function to solve nonlinear least square models (similar to lsqnonlin - Gauss-Newton/Levenberg-Marquardt type of algorithm),
- Addition of the derivative checker,
- Various performance and API improvements.
New interior-point refinement option to improve solution accuracy.
Simplified option-based derivative checker : If the derivative check is successful, KNITRO continues with the optimization (useful during the model development stage).
Deterministic parallel multistart.
Significant speedups for NLPs, MINLPs, and models with range constraints (24% decrease in CPU time vs KNITRO 9.0).
Please note below the upcoming releases of other optimization tools, also distributed by Artelys:
AMPL: Release of the version 9.1 of our comprehensive modeling language for mathematical programming.
FICO Xpress: Release of the version 7.8 of our high performance linear, quadratic and mixed integer programming solver.