Home > Function Evaluation > Function Evaluation Error Limit Exceeded

# Function Evaluation Error Limit Exceeded

ScalingScaling of Intermediate Variables Using the Scale Option in GAMS NLP and DNLP ModelsDNLP Models: What Can Go Wrong? Reformulation from DNLP to NLP Smooth Approximations Are DNLP Models Always Non-smooth? Are NLP Models Always Smooth? APPENDIX A: Algorithmic InformationOverview of GAMS/CONOPT The CONOPT Algorithm Iteration 0: The Initial Point Iteration 1: PreprocessingPreprocessing: Pre-triangular Variables and Constraints Preprocessing: Post-triangular Variables and Constraints Preprocessing: The Optional Crash Procedure Preprocessing: Definitional Equations Iteration 2: Scaling Finding a Feasible Solution: Phase 0 Finding a Feasible Solution: Phase 1 and 2 Linear and Nonlinear Mode: Phase 1 to 4 Linear Mode: The SLP Procedure Linear Mode: The Steepest Edge Procedure Nonlinear Mode: The SQP Procedure How to Select Non-default Options Miscellaneous TopicsTriangular Models Constrained Nonlinear System or Square Systems of Equations Loss of Feasibility Stalling Overflow and NaN (Not A Number) External Equations and Extrinsic Functions APPENDIX B - OptionsAlgorithmic options Debugging options Output options Interface options APPENDIX C: References AuthorArne Drud, ARKI Consulting and Development A/S, Bagsvaerd, Denmark Introduction Nonlinear models created with GAMS must be solved with a nonlinear programming (NLP) algorithm. Currently, there is a large number of different solvers available and the number is growing. The most important distinction between the solvers is whether they attempt to find a local or a global solution. Solvers that attempt to find a global solution (so called Global Solvers) can usually not solve very large models. As a contrast most Local Solvers can work with much larger models, and models with over 10,000 variables and constraints are not unusual. If the model has the right mathematical properties, e.g. is convex, then Local Solvers will find a global optimum. Unfortunately, the mathematical machinery for testing whether a general NLP model is convex or not has not yet been developed (and is expected to be in the class or hard problems). It is almost impossible to predict how difficult it is to solve a particular model with a particular algorithm, especially for NLP models, so GAMS cannot select the best algorithm for you automatically. When GAMS is installed you must select one of the nonlinear programming algorithms as the default solver for NLP mode