This page summarizes how much time is required to solve different
optimization problems. Various solvers are used to compare the
efficiency of the different available approaches.
Methodology
- Run the RobOptim test suite for each solver, the
Hock-Schittkowski Collection problems are included, see
the roboptim-shared-tests
repository to see the actual implementation. The "b" versions
correspond to minor variations of a same problem but using the
RobOptim in a different manner.
- Convert the RobOptim log files, i.e.
journal.log, to a
YAML file using
log_to_yaml.py.
- Tests for which solvers are failing are finally removed from
the dataset (manually for now).
The log file timing information are generated
using Boost.DateTime.
The tests provided here are too small to test sparse solvers in
a robust way. Values for these solvers should be handled with
great care!
The
document defining all the problems is available here.