Papers and Software Download @ Benjamin Inden @ AI Group, Bielefeld University

Papers on Neuroevolution

I have developed the NEATfields method for the design of large artificial neural networks by evolutionary algorithms. The method uses the well known NEAT method for evolving local network structures, but extends this method by adding a number of design patterns for large scale structures that evolution can make use of. The most important of these design patterns is a two-dimensional field of similar neurons.

You can find the technical details of the method in any of the following papers. Each tackles a different set of problems:

Inden, B., Y. Jin, R. Haschke, & H. Ritter. 2012. Evolving neural fields for problems with large input and output spaces. Neural Networks, 28, 24 - 39.

The most detailed description. Problems dealt with: various pattern recognition tasks, control of a multisegmented arm, pole balancing.

Accepted author manuscript

LEGAL NOTICE: this is the author's version of a work that was accepted for publication in Neural Networks. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Neural Networks 28:24-39, 2012. DOI.

Inden, B., Y. Jin, R. Haschke, & H. Ritter. 2011/2012. Evolution of multisensory integration in large neural fields, in: Evolution Artificielle / Lecture Notes in Computer Science 7401, 181-192.

Problems dealt with: Simple models for the integration of input from two modalities for the purpose of using different senses in a cost-optimal way, enhancing the accuracy of classifying noisy input images, or enhancing the spatial accuracy of perception.

Accepted author manuscript

LEGAL NOTICE: The final publication will be available at springerlink.com, Proceedings of Evolution Artificielle, Lecture Notes in Computer Science 7401, Springer-Verlag 2012. DOI: (will be added).

Inden, B., Y. Jin, R. Haschke, & H. Ritter. 2011. Exploiting inherent regularity in control of multilegged robot locomotion by evolving neural fields, in: 2011 Third World Congress on Nature and Biologically Inspired Computing. IEEE Press.

Problems dealt with: locomotion of legged robots with two different morphologies

Accepted author manuscript

LEGAL NOTICE: © 2011 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. DOI

Papers on Coevolutionary Dynamics

I have studied three related number sequence games as simple models of coevolution and demonstrated that they can produce escalating arms races and irreducible functional units of unbounded size under certain conditions. I have also argued that these models show unbounded evolutionary activity according to a previous formal definition. At the moment, I am working to extend these models to various coevolutionary scenarios, and transfer them to the field of evolutionary robotics.

The following paper has been published so far:

Inden, B. 2012. Open-ended Coevolution and the Emergence of Complex Irreducible Functional Units in Iterated Number Sequence Games, in: Proceedings of the 14th annual conference on genetic and evolutionary computation. New York, NY, USA: ACM, 113-120.

Accepted author manuscript

LEGAL NOTICE: © ACM, 2012. This is the author version of the work. It is posted here by permissiomn of ACM for your personal use. Not for redistribution. The definitive version was published in Proceedings of the Genetic and Evolutionary Computation Conference, ISBN: 978-1-4503-1177-9, DOI: 10.1145/2330163.2330180, DOI

Software

The C++-software used for all experiments is closed source. I hope to be able to provide a reimplementation in R within a few months. Please come back to check later. In the meantime, if you have any suggestions for joint scientific work, please contact me.


Papers on Neuroevolution I have developed the NEATfields method for the design of large artificial neural networks by evolutionary algorithms. The method uses the well known NEAT method for evolving local network structures, but extends this method by adding a number of design patterns for large scale structures that evolution can make use of. The most important of these design patterns is a two-dimensional field of similar neurons. You can find the technical details of the method in any of the following papers. Each tackles a different set of problems: Inden, B., Y. Jin, R. Haschke, & H. Ritter. 2012. Evolving neural fields for problems with large input and output spaces. Neural Networks, 28, 24 - 39. The most detailed description. Problems dealt with: various pattern recognition tasks, control of a multisegmented arm, pole balancing. Accepted author manuscript LEGAL NOTICE: this is the author's version of a work that was accepted for publication in Neural Networks. Changes resulting from the publishing process, such as peer review, editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. Changes may have been made to this work since it was submitted for publication. A definitive version was subsequently published in Neural Networks 28:24-39, 2012. DOI. Inden, B., Y. Jin, R. Haschke, & H. Ritter. 2011/2012. Evolution of multisensory integration in large neural fields, in: Evolution Artificielle / Lecture Notes in Computer Science 7401, 181-192. Problems dealt with: Simple models for the integration of input from two modalities for the purpose of using different senses in a cost-optimal way, enhancing the accuracy of classifying noisy input images, or enhancing the spatial accuracy of perception. Accepted author manuscript LEGAL NOTICE: The final publication will be available at springerlink.com, Proceedings of Evolution Artificielle, Lecture Notes in Computer Science 7401, Springer-Verlag 2012. DOI: (will be added). Inden, B., Y. Jin, R. Haschke, & H. Ritter. 2011. Exploiting inherent regularity in control of multilegged robot locomotion by evolving neural fields, in: 2011 Third World Congress on Nature and Biologically Inspired Computing. IEEE Press. Problems dealt with: locomotion of legged robots with two different morphologies Accepted author manuscript LEGAL NOTICE: © 2011 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. DOI

Papers on Coevolutionary Dynamics I have studied three related number sequence games as simple models of coevolution and demonstrated that they can produce escalating arms races and irreducible functional units of unbounded size under certain conditions. I have also argued that these models show unbounded evolutionary activity according to a previous formal definition. At the moment, I am working to extend these models to various coevolutionary scenarios, and transfer them to the field of evolutionary robotics. The following paper has been published so far: Inden, B. 2012. Open-ended Coevolution and the Emergence of Complex Irreducible Functional Units in Iterated Number Sequence Games, in: Proceedings of the 14th annual conference on genetic and evolutionary computation. New York, NY, USA: ACM, 113-120. Accepted author manuscript LEGAL NOTICE: © ACM, 2012. This is the author version of the work. It is posted here by permissiomn of ACM for your personal use. Not for redistribution. The definitive version was published in Proceedings of the Genetic and Evolutionary Computation Conference, ISBN: 978-1-4503-1177-9, DOI: 10.1145/2330163.2330180, DOI

Software The C++-software used for all experiments is closed source. I hope to be able to provide a reimplementation in R within a few months. Please come back to check later. In the meantime, if you have any suggestions for joint scientific work, please contact me.