ScalarSolvers

Collection of different algorithms to minimize problems like the one dimensionnal problems in OptimizationProblems.

How to use

With NLPModels, OptimizationProblems, State and Stopping. Here is a simple example:

nlp = MathProgNLPModel(AMPGO02());
nlp_at_x = NLPAtX([3.7])
nlp_stop = NLPStopping(nlp, Stopping.unconstrained, nlp_at_x)
optimal, stop = TR_Nwt(nlp, nlp_stop, verbose = true)

The output should be:

 iter  t         gₖ          Δ        pred         ared
    0 2.70e+00  -3.94e+00  1.00e+00
    1 3.70e+00  2.40e+00  1.00e+00 -6.44e+00 -1.60e+00
    2 2.93e+00  -4.13e+00  1.00e+00 -9.27e-01 6.51e-01
    3 2.93e+00  -4.13e+00  5.00e-01 -2.44e+00 -9.85e-02
    4 3.43e+00  4.16e-01  5.00e-01 -1.64e+00 -1.08e+00
    5 3.39e+00  -1.02e-02  1.00e+00 -8.33e-03 -8.18e-03
    6 3.39e+00  -4.32e-06  2.00e+00 -4.83e-06 -4.83e-06
    7 3.39e+00  -7.75e-13  4.00e+00 -8.56e-13 -8.56e-13

Installing

julia> add https://github.com/Goysa2/ScalarSolvers

Line Search applications

All the algorithms presented in this packages are adapted for Line Search in the package LineSearch. It is possible to directly use these algorithm for line search using an appropriate stopping criterion with a Stopping object.

References

All the algorithms can be found here:

J. Nocedal and S.Wright Numerical Optimization

J.P Dusseault Univariate diffentiable optiization algorithms and LineSearch computation