Numerical Algorithms is offering step-by-step instructions on how researchers can select the best suited local or global optimisation routines by making use of the Decision Trees for optimisation.
The numerical software development organisation devotes significant research and development resources to continually advance methods for solving optimisation problems and similar computational challenges.
The NAG Decision Trees are part of the documentation for one of the most rigorously tested and documented sets of optimisation routines and other mathematical and statistical algorithms in the world, collected into the NAG Library.
The NAG Library of routines, including the optimisation chapters, can be called from diverse environments such as C++, Fortran, Matlab and R.
Many of those who use NAG's routines as the building blocks of their applications rely on the knowledge base in NAG's exhaustive documentation as a part of the 'future-proofing' of their application development investments.
The Decision Trees, which are a feature of this documentation, are especially useful in helping both new and experienced users to select the appropriate routine for the problem at hand in minutes.
Dr David Sayers, a principal technical consultant at NAG, said: 'For maximum efficiency, different algorithms should be used for a different problem types.
'Often these types are characterised by the type of objective function - that is to be minimised or maximised - and by the types of constraints that are to be applied.
'Objective functions might be linear, quadratic (positive-definite or indefinite) or nonlinear.
'They may have a special form, like a sum of squares.
'They may be sparse or dense and they may be smooth or discontinuous.
'Combine these with the options for constraints: none, simple bound, linear or genuine nonlinear and we can see that a comprehensive chapter of optimisation routines can be very large.
'To help the user to choose the right routine decision trees are invaluable.'
