wiki:WikiStart

oBB

oBB is an algorithm for the parallel global optimization of functions with Lipchitz continuous gradient or Hessian.

The algorithm is described in the paper:

Cartis, C., Fowkes, J. M., and Gould, N. I. M. (2015). Branching and Bounding Improvements for Global Optimization Algorithms with Lipschitz Continuity Properties. Journal of Global Optimization, vol. 61, no. 3, pp. 429–457.

The latest version contains an optional range reduction strategy that improves performance but may not always guarantee global optimality. For details please see the Master's thesis:

Guida, A. (2015). A Branch and Bound Algorithm for the Global Optimization and its Improvements. Master's Thesis, Faculty of Engineering, University of Florence.

Documentation and Source Code

Comprehensive HTML documentation is available at http://packages.python.org/oBB. Source code for oBB is hosted on GitHub: https://github.com/coin-or/oBB

Requirements

oBB requires the following software to be installed:

Additionally, the following python packages should be installed (these will be installed automatically if using pip, see Installation using pip):

Optionally, matplotlib 1.1.0 or higher (http://www.matplotlib.org/) may be manually installed for visualising the algorithm in 2D.

Installation using pip

For easy installation, use pip (http://www.pip-installer.org/) as root:

$ [sudo] pip install --pre obb

or alternatively easy_install:

$ [sudo] easy_install obb

If you do not have root privileges or you want to install oBB for your private use, you can use:

$ pip install --pre --user obb

which will install oBB in your home directory.

Note that if an older install of oBB is present on your system you can use:

$ [sudo] pip install --pre --upgrade obb

to upgrade oBB to the latest version.

Manual installation

Alternatively, you can download the source code and unpack as follows:

$ wget http://pypi.io/packages/source/o/oBB/oBB-0.8b.zip
$ unzip oBB-0.8b.zip
$ cd oBB-0.8b

and then build and install manually using:

$ python setup.py build
$ [sudo] python setup.py install

If you do not have root privileges or you want to install oBB for your private use, you can use:

$ python setup.py install --user

instead.

Testing

oBB includes a command line test script to check that the installation was successfull. To run the test simply type the following into your shell:

$ test_obb

This will run oBB using MPI on one processor core for a simple 2D sum of sines problem.

Note that if using the MPICH implementation of MPI you first need to start an mpd daemon in the background:

$ mpd &

but this is not necessary for other MPI implementations, e.g. OpenMPI.

Uninstallation

If oBB was installed using pip you can uninstall as follows:

$ [sudo] pip uninstall obb

If oBB was installed manually you have to remove the installed files by hand (located in your python site-packages directory).

Reporting Bugs

Please use GitHub's issue tracker or the COIN-OR ticket system to report bugs (note that you must be registered and logged in first). To ensure that your ticket is dealt with as soon as possible, please try to be as exhaustive as you can in the bug report.

License

This algorithm is released under the GNU LGPLv3 license.

Last modified 11 months ago Last modified on Aug 2, 2016 5:37:34 AM