Numerical optimization problems occur almost everywhere - in physics, economics, chemistry, biology etc. For example:
- planes, helicopters, vehicles fuel consumption minimizing
- shop profits maximizing
- seedlings of crops growth (in greenhouses) maximizing
- gaskets ways communications minimizing (auto and transportation industries, the Internet network)
- with various kinds of restrictions, such as
- minimum and maximum speed
- working capital, the risk for devastation, the showcase
- illumination, water and financial costs on them
- capacity of branches and nodes
Software for solving numerical optimization costs thousands, and sometimes tens of thousands $ (example and another one)
In addition, you should spend approximately 10% on updating software libraries annually.
In recent times, due to sponsorship from various funds, universities, some corporations (IBM), lots of viable free numerical optimization software have appeared. However, most packages have license GPL/LGPL containing copyleft restriction, that prevents using of the software by closed or any other non-free code. Also, most of them is C- or Fortran-written, that doesn't allow RAD (Rapid Application Development) and has lots of cross-platform troubles (Linux, Windows, MacOS etc).
Our 100%-free OpenOpt (Python-written, license: BSD) consists of few own solvers (e.g. ralg, interalg, gsubg) and includes interfaces to a number of other (mostly free as well), many of which are written in C/Fortran (BTW, for matrix operations OpenOpt uses C-written NumPy, that involves Intel MKL, AMD ACML and is BSD-licensed as well). Involving FuncDesigner can yield additional RAD abilities and Automatic differentiation for some classes of NonLinearProblems. All the above mentioned forced OpenOpt to take high positions in google. For now (September 2010) we have already 120...150 visitors daily, ~15% of them visit installation webpage, and many others install our soft from PYPI, mloss.org, pythonxy.com, Debian, Alt Linux, SAGE, mac.softpedia.com, macosforge.org, darwinports.com repositories etc.
As many experts note, software development in Python is approximately 2 times faster than Fortran one (+ requires much less code lines amount), and it should be noted that this is not just programmer man-hours, but other staff as well, rental payments etc. Note: connecting C or Fortran code to Python is much easier than, for example, using MATLAB MEX-functions. Thus, even MIT (lead USA technical institute) has begone to study Python (as well as several our Ukrainian universities).
|Made by Dmitrey|