Introduction

JavaEvA is a framework for Evolutionary and general Optimization Algorithms implemented in Java. It is developed as resumption of the former Evolutionary Algorithms (EvA) software package. Using the Java mechanism of Remote Method Invocation (RMI) the algorithms of JavaEvA can be distributed via Intra- or Internet. Thus, it is possible to handle complex optimisation problems. JavaEvA as a client server application is divided into two parts: a swing-based graphical user interface (EvAClient) and the optimization kernel (EvAServer) which contains the different optimization algorithms.

JavaEvA includes Evolutionary Algorithms (EA) like Genetic Algorithms (GA), Evolutionary Strategies (ES), Population Based Incremental Learning (PBIL) and many more. Besides, model-based Optimisation Techniques like methods from the research field of Design of Experiment, Efficient Global Optimisation (EGO) and Model-Assisted Evolution Strategies (MAES) are implemented in JavaEvA. Concerning model-based Optimisation Techniques, several mathematical models can be used like Gaussian Processes, Support Vector Machines, Radial-Basis-Function networks etc.. A comprehensive library of neural network implementations is available, using the JavaNNS package.

JavaEvA was developed as part of the BMBF-Project called "Automated Crystallization". JavaEvA will be used here to optimise parameters for an automated production process for crystalline catalysts that yield the highest production output and the best catalytic properties of the produced crystals.

JavaEvA has been applied as teaching aid in lecture tutorials, on several optimisation problems within our research group and as developing platform in student research projects.