https://github.com/odiroot/pyho

Hybrid Optimizer in Python (Github mirror)
https://github.com/odiroot/pyho

Last synced: 23 days ago
JSON representation

Hybrid Optimizer in Python (Github mirror)

• Host: GitHub
• URL: https://github.com/odiroot/pyho
• Owner: odiroot
• Created: 2012-06-03T12:26:03.000Z (over 12 years ago)
• Default Branch: master
• Last Pushed: 2012-06-05T21:07:33.000Z (over 12 years ago)
• Last Synced: 2024-04-14T19:33:03.353Z (7 months ago)
• Language: C
• Homepage: https://bitbucket.org/odiroot/pyho/
• Size: 5.05 MB
• Stars: 1
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.txt

Awesome Lists containing this project

README

        

=================================

PyHO - Hybrid optimizer in Python

=================================

PyHO is a prototype of an universal two-stage optimizer written in Python.

This project is based on the original C++ optimizer module from `FMDT`

(Flow Meter Design Tool) project.

Current status

--------------

- Objective function optimization with a genetic algorithm engine

    - Support for customized genetic operators

    - Support for `PyEvolve` specific `allele` operators

- Second stage optimization with Levenberg-Marquardt algorithm

- Optimization speedup using memoization

- Parallel processing using network or local worker processes

    - `Local mode` with automatic worker management

    - `Network mode` with arbitrary number of worker nodes

- Currently optimizer and worker processes need access to the same file system

  (passing files through the network channel is not implemented nor designed)

- Reading original FMDT input file formats (coil and grid definitions)

- Optimizer can be successfully launched by the `FMDT` GUI.

Dependencies

------------

**PyHO has some hard dependencies**:

* ``zmq`` Python package (called ``python-zmq`` in Debian based systems) -

  version 2.1.9 or newer is highly recommended

* The `ZeroMQ` library (or ``libzmq1``) with sources (``libzmq-dev``) - again

  the 2.1.9+ version is preferred.

If you want to use hybrid optimization mode make sure you also have:

* Source distribution of BLAS (``libblas-dev``) and LAPACK (``liblapack-dev``) libraries

* The `NumPy` Python library (try ``python-numpy``)

PyHO also uses following Python packages:

* ``Cython`` (version at least 0.16)

* ``pyximport``

* ``PyEvolve``

* ``levmar`` (modified version supporting user initiated breaks)

All these libraries are included in the ``libs`` directory for convenience.

When present in the standard Python path, system installed versions take

precedence.

**A note of warning**: This project uses custom ``levmar`` package, officially

available version cannot be used, as it will break during hybrid optimization.

Quick start

-----------

You can start the optimizer directly by launching ``run_optimizer`` Python

script from the root project directory. Running it with ``--help`` option

will show all available options.

You can use example evaluator based on FMDT project by using ``run_bc_evaluator``

shell script. There are example coil/grid definition files located in ``assets``

directory under ``examples/fmdt/`` path.

There are 32-bit Linux and 64-bit Linux precompiled binaries of `FMDT` GUI

application packaged in a ``GUI`` directory. Run it by calling ``Kanal_v1``

shell script.