Assembler Encoding with Evolvable Operations
Institute of Naval Weapon, Polish Naval Academy
81-103 Gdynia, ul. Śmidowicza 69
E-mail: t.praczyk@amw.gdynia.pl
Received:
Received: 23 June 2015; revised: 25 August 2015; accepted: 27 August 2015; published online: 04 September 2015
DOI: 10.12921/cmst.2015.21.03.004
Abstract:
Assembler Encoding is a neuro-evolutionary method which represents a neural network in the form of a linear program. The program consists of operations and data and its goal is to produce a matrix including all the information necessary to construct a network. In order for the programs to produce effective networks, evolutionary techniques are used. A genetic algorithm determines an arrangement of the operations and data in the program and parameters of the operations. Implementations of the operations do not evolve, they are defined in advance by a designer. Since operations with predefined implementations could narrow down applicability of Assembler Encoding to a restricted class of problems, the method has been modified by applying evolvable operations. To verify effectiveness of the new method, experiments on the predator-prey problem were carried out. In the experiments, the task of neural networks was to control a team of underwater-vehicles-predators whose common goal was to capture an underwater-vehicle-prey behaving by a simple deterministic strategy. The paper describes the modified method and reports the experiments.
Key words:
References:
[1] T. Aaltonen,et al.,Measurement of the topquark mass with
dilepton events selected using neuroevolution at CDF, Phys-
icalReviewLetters (2009).
[2] B. Allen, P.Faloutsos, Complex networks of simple neurons
for bipedal locomotion, In IEEE/RSJ International Confer-
enceon Intelligent Robots and Systems (2009).
[3] A. Cangelosi, D. Parisi, S. Nolfi, Cell divisionand migration
in a genotype for neural networks,Network: computation in
neuralsystems 5(4), 497-515, (1994).
[4] J. Gauci,K. Stanley, Generating large-scale neural networks
through discovering geometric regularities,In: Proceedings of
the Genetic and Evolutionary Computation Conference,pp.
997-1004 (2007).
[5] D. E. Goldberg, Genetic algorithms in search, optimization
and machine learning, Addison Wesley, Reading, Massachusetts,1989.
[6] F. Gruau, Neural network Synthesis Using Cellular Encoding
And The Genetic Algorithm, PhD Thesis, Ecole Normale
Superieurede Lyon1994.
[7] H. Kitano, Designing neural networks using genetic algo-
rithms with graph generation system, Complex Systems 4,
461-476,(1990).
[8] K. Krawiec, B. Bhanu, Visual Learning by Coevolutionary
Feature Synthesis, IEEE Trans. on Systems, Man, and Cy-
bernetics, PartB: Cybernetics 35, 409-425 (2005).
[9] T. Kubaty, L. Rowinski, Mine counter vehicles for Baltic Navy,
Internet, http://www.underwater.pg.gda.pl.
[10] J. Lehman, K. O. Stanley, AbandoningObjectives: Evolution
through the SearchforNovelty Alone,Evolutionary Compu-
tation 19, 189-223 (2011).
[11] S.Luke and L. Spector, Evolving Graphs andNetworks with
Edge Encoding: Preliminary Report, In John R. Koza, ed.,
Late Breaking Papers at the Genetic Programming 1996 Conference,
(Stanford University,CA,USA,Stanford Bookstore,
1996) 117-124.
[12] G.F. Miller, P.M. Todd, S.U. Hegde, Designing Neural Net-
works Using Genetic Algorithms, Proceedings of the Third
International Conference on Genetic Algorithms. 379-384. of
Schaffer J.D. (1989).
[13] D. E.Moriarty, Symbiotic Evolution ofNeural Networks in
Sequential Decision Tasks, PhD thesis, The University of
Texas atAustin,TR UT-AI97-257,(1997).
[14] S. Nolfi, D.Parisi, Growing neural networks,InC. G. Lang-
ton, ed.,Artificial Life III, Addison-Wesley, (1992).
[15] P. Nordin, W. Banzhaf, F. Francone, Efficient Evolution of
Machine Code for CISC Architectures usingBlocks and Homologous Crossover,
Advances in Genetic Programming III,
MITPress, L. Spectorand W. Langdon and U.O’Reilly and
P. Angeline, pages. 275-299 (1999).
[16] M. Potter, The Design and Analysis of a Computational
Model of Cooperative Coevolution, PhDthesis, George Ma-
son University, Fairfax, Virginia(1997).
[17] M. A. Potter, K. A. De Jong, Cooperative coevolution: An
architecture for evolving coadapted subcomponents, Evolu-
tionary Computation 8(1), 1-29 (2000).
[18] T.Praczyk, Modular networks in Assembler Encoding, Com-
putational Methods in Science and Technology 14(1),27-38
(2008).
[19] T.Praczyk, Using assembler encoding to solve inverted pen-
dulum problem, Computing and Informatics 28, 895-912
(2009).
[20] T. Praczyk, Forming NeuralNetworks by Means ofAssem-
bler Encoding, Intelligent Automation and Soft Computing
17(3), 319-331 (2011).
[21] T. Praczyk, P. Szymak, Decision System for a Team of Au-
tonomousUnderwaterVehicles- Preliminary Report, Neuro-
computing, doi:10.1016/j.neucom.2011.05.013.
[22] T.Praczyk, Solving the pole balancing problem by means of
Assembler Encoding, Journal of Intelligent and Fuzzy Sys-
tems 26, 857-868 (2014).
[23] T. Praczyk, Diverse neural architectures in Assembler En-
coding, Computational Methods in Science and Technology
20(1), 21-34, (2014).
[24] K. O. Stanley, R. Miikkulainen, Evolving neural networks
through augmenting topologies,Evolutionary Computation
10, 99-127, (2002).
[25] K. O. Stanley, R. Miikkulainen, Competitive coevolution
through evolutionary complexification, Journal of Artificial
Intelligence Research 21, 63–100 (2004).
Assembler Encoding is a neuro-evolutionary method which represents a neural network in the form of a linear program. The program consists of operations and data and its goal is to produce a matrix including all the information necessary to construct a network. In order for the programs to produce effective networks, evolutionary techniques are used. A genetic algorithm determines an arrangement of the operations and data in the program and parameters of the operations. Implementations of the operations do not evolve, they are defined in advance by a designer. Since operations with predefined implementations could narrow down applicability of Assembler Encoding to a restricted class of problems, the method has been modified by applying evolvable operations. To verify effectiveness of the new method, experiments on the predator-prey problem were carried out. In the experiments, the task of neural networks was to control a team of underwater-vehicles-predators whose common goal was to capture an underwater-vehicle-prey behaving by a simple deterministic strategy. The paper describes the modified method and reports the experiments.
Key words:
References:
[1] T. Aaltonen,et al.,Measurement of the topquark mass with
dilepton events selected using neuroevolution at CDF, Phys-
icalReviewLetters (2009).
[2] B. Allen, P.Faloutsos, Complex networks of simple neurons
for bipedal locomotion, In IEEE/RSJ International Confer-
enceon Intelligent Robots and Systems (2009).
[3] A. Cangelosi, D. Parisi, S. Nolfi, Cell divisionand migration
in a genotype for neural networks,Network: computation in
neuralsystems 5(4), 497-515, (1994).
[4] J. Gauci,K. Stanley, Generating large-scale neural networks
through discovering geometric regularities,In: Proceedings of
the Genetic and Evolutionary Computation Conference,pp.
997-1004 (2007).
[5] D. E. Goldberg, Genetic algorithms in search, optimization
and machine learning, Addison Wesley, Reading, Massachusetts,1989.
[6] F. Gruau, Neural network Synthesis Using Cellular Encoding
And The Genetic Algorithm, PhD Thesis, Ecole Normale
Superieurede Lyon1994.
[7] H. Kitano, Designing neural networks using genetic algo-
rithms with graph generation system, Complex Systems 4,
461-476,(1990).
[8] K. Krawiec, B. Bhanu, Visual Learning by Coevolutionary
Feature Synthesis, IEEE Trans. on Systems, Man, and Cy-
bernetics, PartB: Cybernetics 35, 409-425 (2005).
[9] T. Kubaty, L. Rowinski, Mine counter vehicles for Baltic Navy,
Internet, http://www.underwater.pg.gda.pl.
[10] J. Lehman, K. O. Stanley, AbandoningObjectives: Evolution
through the SearchforNovelty Alone,Evolutionary Compu-
tation 19, 189-223 (2011).
[11] S.Luke and L. Spector, Evolving Graphs andNetworks with
Edge Encoding: Preliminary Report, In John R. Koza, ed.,
Late Breaking Papers at the Genetic Programming 1996 Conference,
(Stanford University,CA,USA,Stanford Bookstore,
1996) 117-124.
[12] G.F. Miller, P.M. Todd, S.U. Hegde, Designing Neural Net-
works Using Genetic Algorithms, Proceedings of the Third
International Conference on Genetic Algorithms. 379-384. of
Schaffer J.D. (1989).
[13] D. E.Moriarty, Symbiotic Evolution ofNeural Networks in
Sequential Decision Tasks, PhD thesis, The University of
Texas atAustin,TR UT-AI97-257,(1997).
[14] S. Nolfi, D.Parisi, Growing neural networks,InC. G. Lang-
ton, ed.,Artificial Life III, Addison-Wesley, (1992).
[15] P. Nordin, W. Banzhaf, F. Francone, Efficient Evolution of
Machine Code for CISC Architectures usingBlocks and Homologous Crossover,
Advances in Genetic Programming III,
MITPress, L. Spectorand W. Langdon and U.O’Reilly and
P. Angeline, pages. 275-299 (1999).
[16] M. Potter, The Design and Analysis of a Computational
Model of Cooperative Coevolution, PhDthesis, George Ma-
son University, Fairfax, Virginia(1997).
[17] M. A. Potter, K. A. De Jong, Cooperative coevolution: An
architecture for evolving coadapted subcomponents, Evolu-
tionary Computation 8(1), 1-29 (2000).
[18] T.Praczyk, Modular networks in Assembler Encoding, Com-
putational Methods in Science and Technology 14(1),27-38
(2008).
[19] T.Praczyk, Using assembler encoding to solve inverted pen-
dulum problem, Computing and Informatics 28, 895-912
(2009).
[20] T. Praczyk, Forming NeuralNetworks by Means ofAssem-
bler Encoding, Intelligent Automation and Soft Computing
17(3), 319-331 (2011).
[21] T. Praczyk, P. Szymak, Decision System for a Team of Au-
tonomousUnderwaterVehicles- Preliminary Report, Neuro-
computing, doi:10.1016/j.neucom.2011.05.013.
[22] T.Praczyk, Solving the pole balancing problem by means of
Assembler Encoding, Journal of Intelligent and Fuzzy Sys-
tems 26, 857-868 (2014).
[23] T. Praczyk, Diverse neural architectures in Assembler En-
coding, Computational Methods in Science and Technology
20(1), 21-34, (2014).
[24] K. O. Stanley, R. Miikkulainen, Evolving neural networks
through augmenting topologies,Evolutionary Computation
10, 99-127, (2002).
[25] K. O. Stanley, R. Miikkulainen, Competitive coevolution
through evolutionary complexification, Journal of Artificial
Intelligence Research 21, 63–100 (2004).
