Benchmarking Nonlinear Constrained Optimization algorithms
This is work in progress
There are two types of problems: fixed problems and scalable ones. A concrete problem inherites from the base classe base.ConstrainedTestProblem
.
A fixed problem has fixed dimension dim
and number of inequality constraints m
. Most of its methods are static hence it can be used without instantiation. However the property m
cannot be accessed in this class. Later improvement is to define class properties.
A scalable problem needs instantiation with dimension and number of constraints.
- Add the decorators
arrayize
andrealfunction
everywhere. - Implement the gradient everywhere it is possible
Just do pip install .
in this repository, everything relies on numpy
and scipy.optimize
if you want to run experiments.