Winkler, Jonas; Hauke, Sascha; Pyka, Martin; Heider, Dominik:

JACVANN: A Java Framework For Complex Valued Artificial Neural Networks

In: Systemics and Informatics World Network, Jg. 10 (2010), S. 48-55
ISSN: 2044-7272
Zeitschriftenaufsatz / Fach: Informatik; Biologie
Abstract:
JACVANN (Java Framework for Complex Valued Artificial Neural Networks) is an open source framework for complex-valued artificial neural networks. It provides several initialization procedures, such as the Nguyen-Widrow method, learning algorithms, such as backpropagation and resilient propagation, as well as different evaluation methods for complex-valued classification problems. Furthermore, we implemented a procedure for transferring real-valued problems into complex space to harness the strengths of our framework. The framework was tested with several geometric and classification problems from the literature. It is shown that JACVANN is both robust and flexible due to its modular composition.
Availability: http://www.uni-due.de/~hy0546/JACVANN/
Contact: dominik.heider@uni-due.de
Requirements: Java 5.0 or higher