NeuroGraph - A Simulation Environment for Neural Networks, Genetic Algorithms and Fuzzy Logic

Abstract

Neuro Graph is a simulation environment for neural networks. It provides an easy to use graphical user interface to design, construct and execute neural networks. The most important design goals were easy extensibility, good performance and the ability to solve real world problems. Extensibility is ensured by an extremely flexible, hierarchical internal representation (data structure), which can be easily manipulated by graphical tools. Current available extensions to the neural network simulator are components for genetic algorithms and fuzzy logic. As examples the automated generation of a neural network by a genetic algorithm and a fuzzy controller are shown. This data structure can be interpreted for debugging purposes or can be executed directly for high performance computing. In the later case C or C++ code is generated from the internal representation of the neural, gentic or fuzzy functions, compiled and stored in an executable process (library). Neuro Graph has a file or process interface to interconnect which other processes. It is also possible to generate C or C++ code representing the neural network, genetic algorithm or fuzzy component which can be incorporated in other applications.