EvoFrame:

EvoFrame is to EVOLUTION STRATEGIEs what MicroGA is to GENETIC ALGORITHMs, a toolkit for application development incorporating ESs as the OPTIMIZATION engine.

EvoFrame is an object oriented implemented programming tool for evolution strategies (Rechenberg/Schwefel, Germany) for easy implementation and solution of numerical and combinatorical problems. EvoFrame gives you freedom of implementing every byte of the optimization principle and its user interface. You can focus on the optimization problem and forget about all the rest.

EvoFrame is available as Version 2.0 in Borland-Pascal 7.0 and Turbo-Vision for PC's and as Version 1.0 in C++ for Apple Macintosh using MPW and MacApp. Both implementations allow full typed implementation, i.e. no more translation from problem specific format to an optimization specific one. A prototyping tool (cf REALizer) exists for both platforms too.

EvoFrame allows pseudoparallel optimization of many problems at once and you can switch optimization parameters and internal methods (i.e. quality function etc.) during runtime and during optimization cycle. Both tools can be modified or extended by overloading existing methods for experimental use. They are developed continously in correlation to new research results.

The PC version is prepared for experimental use due to a comprehensive protocolling mechanism of optimzation cycles and user data. It also allows compilation of executable files with different complexity by setting conditional compilation flags. It can be used with 3 levels of stacked POPULATIONs.

The Mac version is the more complex (recursive) implementation. It allows stacking of any number of populations for modelling of complex systems. Theory stops at multipopulation level at the time. EvoFrame for Mac is ready for the future, allowing any number of population levels.

Ask for porting the Mac version (C++) to any other platform, i.e. X Windows.

REALizer is a tool for rapid prototyping of EvoFrame applications. It's an override of the corresponding framework which is prepared to optimize using a vector of real numbers. All methods for standard EVOLUTION and file handling, etc. are ready implemented. The remaining work for the user is to define a constant for the problem size, fill in the quality function and start the optimization process.

For further information, current prices and orders, contact: Wolfram Stebel, Optimum Software, Braunfelser Str. 26, 35578 Wetzlar, Germany. Net: <optimum@applelink.apple.com>