EASY

EASY is a general purpose, high-fidelity software for the search of optimal solutions in single- or multi-objective problems through Generalized Evolutionary Algorithms.

The employed search algorithm does not necessarily coincide with genetic algorithms or evolution strategies; it is just a matter of appropriate parameter setting to switch to the aforementioned well-known techniques. EASY operates through handling and evolving three sets of populations, namely the parent, offspring and the elite or archival set of solutions. In MOO problems, the latter is used for the storage of the Pareto front of optimal solutions.

The use of EASY is easy, even for an unexperienced user. The user should define lower and upper bounds of the design variables along with the evolution parameters and specify the evaluation software (as a sequence of names of executable files) in script task.bat. The communication of EASY with the user's evaluation software is automatically carried out through two files: task.dat communicates any 'current' set of values of the design variables to the evaluation tools defined in task.bat; this performs the necessary evaluation(s) and returns the objective function value(s) in task.res. The failure of evaluation of any individual returns less objective values than expected and this is taken into account by EASY, without user interventions.

The strong point of EASY is the optional coupling of EAs with computational intelligence techniques, in order to minimize the CPU cost of the overall optimization. The implementation mode is based on the concept of the so-called Inexact Pre-Evaluations (IPE). Low-cost, local metamodels (artificial neural networks, in particular) are on-line trained for each and every new candidate solution. A number of neighboring, already evaluated solutions is located and an artificial neural network is trained on them. By doing so, the network may act as a cheap surrogate evaluation tool. In each generation, the entire population is first evaluated using local metamodels. Then, the most promising individuals among them are identified and re-evaluated through the exact and costly evaluation tool.

EASY is currently available for Microsoft Windows, Unix and Linux operating systems. The user interface is built on the Java platform while the core evolutionary system is programmed in C++ for high performance. EASY versions for other computing platforms can be (or have already been) made available upon request.

more at EASY site