Logic Journal of IGPL Advance Access published online on September 3, 2009
Logic Journal of IGPL, doi:10.1093/jigpal/jzp043
Structure optimization of reservoir networks
Institut für Neuroinformatik, Ruhr-Universität Bochum, 44780 Bochum, Germany.
E-mail: Benjamin.Roeschies{at}neuroinformatik.rub.de; Christian.Igel{at}neuroinformatik.rub.de
Reservoir computing originally relies on a random, fixed population of recurrently connected neurons, but in practice there is a need for algorithms that tailor this reservoir to problem classes. We propose an evolutionary algorithm to adapt size and topology of the network as well as synaptic strengths of connections within the reservoir. Experiments show that the evolved networks significantly outperform standard architectures in terms of prediction performance and compare well with the results of alternative neuroevolution approaches. For applications in which both space and execution time complexity are important, we present an evolutionary multi-objective algorithm addressing the trade-off between identifying dynamical systems with maximum accuracy and minimizing reservoir complexity. This vector optimization algorithm utilizes the contributed hypervolume for selection and for online adaptation of operator probabilities, which considerably improves the search process. Analysis of the evolved networks reveals insights about task-dependent macroscopic properties of the adapted reservoirs.
Key Words: recurrent neural networks • reservoir networks • evolutionary computation • time series prediction • multiobjective optimization
References
-
[1] Angeline PJ, Saunders GM, Pollack JB. An evolutionary algorithm that constructs recurrent neural networks. IEEE Transactions on Neural Networks (1994) 5(1):54–65.[CrossRef][Web of Science][Medline]
[2] Beume N, Naujoks B, Emmerich M. SMS-EMOA: Multiobjective selection based on dominated hypervolume. European Journal of Operational Research (2007) 181(3):1653–1669.[CrossRef][Web of Science]
[3] Davis L. Handbook of Genetic Algorithms (1991) New York: Van Nostrand Reinhold.
[4] Deb K, Agrawal S, Pratap A, Meyarivan T. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation (2002) 6(2):182–197.[CrossRef][Web of Science]
[5] Floreano D, Dürr P, Mattiussi C. Neuroevolution: from architectures to learning. Evolutionary Intelligence (2008) 1(1):47–62.[CrossRef]
[6] Fogel DB. Evolutionary computation: toward a new philosophy of machine intelligence (2006) 2 edition. IEEE Press.
[7] Fonseca CM, Fleming PJ. On the performance assessment and comparison of stochastic multiobjective optimizers. In: Parallel Problem Solving from Nature (PPSN IV)—Voigt Hans-Michael, Ebeling Werner, Rechenberg Ingo, Schwefel Hans-Paul, eds. (1996) Springer-Verlag. 584–593. LNCS.
[8] Gomez F, Miikkulainen R. Incremental evolution of complex general behavior. Adaptive Behavior (1997) 5(3-4):317.
[9] Gomez F, Schmidhuber J, Miikkulainen R. Accelerated neural evolution through cooperatively coevolved synapses. Journal of Machine Learning Research (2008) 9:937–965.[Web of Science]
[10] Grunert da Fonseca V, Fonseca CM, Hall AO. Inferential performance assessment of stochastic optimisers and the attainment function. In: First International Conference on Evolutionary Multi-Criterion Optimization (EMO)—Zitzler E, Deb K, Thiele L, Coello Coello CA, Corne D, eds. (2001) Springer-Verlag. 213–225. volume 1993 of LNCS.
[11] Hansen N, Ostermeier A. Completely Derandomized Self-Adaptation in Evolution Strategies. Evolutionary Computation (2001) 9(2):159–195.[CrossRef][Web of Science][Medline]
[12] Heidrich-Meisner V, Igel C. Neuroevolution strategies for episodic reinforcement learning. Journal of Algorithms (2009) doi:10.1016/j.jalgor.2009.04.002.
[13] Hochreiter S, Schmidhuber J. Long short-term memory. Neural Computation (1997) 9(8):1735–1780.[CrossRef][Web of Science][Medline]
[14] Igel C, Friedrichs F, Wiegand S. Evolutionary optimization of neural systems: The use of strategy adaption. In: Trends and Applications in Constructive Approximation—de Bruin MG, Mache DH, Szabados J, eds. (2005) 151. Birkhäuser Verlag. 103–123.[CrossRef]
[15] Igel C, Glasmachers T, Heidrich-Meisner V. Shark. Journal of Machine Learning Research (2008) 9:993–996.[Web of Science]
[16] Igel C, Hansen N, Roth S. Covariance matrix adaptation for multi-objective optimization. Evolutionary Computation (2007) 15(1):1–28.[CrossRef][Web of Science][Medline]
[17] Igel C, Kreutz M. Operator adaptation in evolutionary computation and its application to structure optimization of neural netzworks. Neurocomputing (2003) 55(1–2):347–361.[CrossRef][Web of Science]
[18] Igel C, Sendhoff B. Genesis of organic computing systems: Coupling evolution and learning. In: Organic Computing—Würtz R, ed. (2008) Springer-Verlag. 141–166. chapter 7.
[19] Igel C, Toussaint M. Neutrality and self-adaptation. Natural Computing (2003) 2(2):117–132.[CrossRef]
[20] Ishii K, van der Zant T, Be
anovi
V, Plöger P. Identification of motion with echo state network. In: Oceans’04. MTS/IEEE Techno-Ocean’04 (2004) 3:1205–1210.
[21] Jaeger H. The ‘‘echo state’’ approach to analysing and training recurrent neural networks. (2001) German National Research Center for Information Technology. GMD Report 148.
[22] Jaeger H. Echo state network. Scholarpedia (2007) 2(9):2330.
[23] Jaeger H, Haas H. Harnessing nonlinearity: Predicting chaotic systems and saving energy in wireless communication. Science (2004) 304:78–80.
[24] Jaeger H, Lukosevicius M, Popovici D, Siewert U. Optimization and applications of echo state networks with leaky integrator neurons. Neural Networks (2007) 20(3):335–352.[CrossRef][Web of Science][Medline]
[25] Jiang F, Berry H, Schoenauer M. Supervised and evolutionary learning of echo state networks. In: Parallel Problem Solving from Nature (PPSN X)—Rudolph G, ed. (2008) Springer-Verlag. 215–224. volume 5199 of LNCS.
[26] Jin Y. Multi-objective Machine Learning (2006) Springer-Verlag.
[27] Jin Y, Sendhoff B. Pareto-based multiobjective machine learning: An overview and case studies. IEEE Transactions on Systems, Man, and Cybernetics, Part C: Applications and Reviews (2008) 38(3):397–415.[Web of Science]
[28] Marder E, Goaillard J-M. Variability, compensation and homeostasis in neuron and network function. Nature Reviews Neuroscience (2006) 7:563–574.[CrossRef][Web of Science][Medline]
[29] Moriarty DE, Miikkulainen R. Efficient reinforcement learning through symbiotic evolution. Machine Learning (1996) 22:11–32.[Web of Science]
[30] Natschläger T, Maass W, Markram H. The liquid computer - a novel strategy for real-time computing on time series. Special Issue on Foundations of Information Processing of TELEMATIK (2002) 8(1):39–43.
[31] Pellecchia A, Igel C, Edelbrunner J, Schöner G. Making driver modeling attractive. IEEE Intelligent Systems (2005) 20(2):8–12.[CrossRef]
[32] Reed RD, Marks RJ. Neural Smithing (1999) MIT Press.
[33] Schmidhuber J, Gagliolo M, Wierstra D, Gomez F. Evolino for recurrent support vector machines. (2006) Evere, Belgium: d-side publications. 593–598. In 14th European Symposium on Artificial Neural Networks (ESANN 2006).
[34] Schmidhuber J, Wiestra D, Gagliolo M, Gomez F. Training recurrent networks by evolino. Neural Computation (2007) 19(3):757–779.[CrossRef][Web of Science][Medline]
[35] Schrauwen B, Verstraeten D, Van Campenhout J. Adapting reservoirs to get gaussian distributions. Verleysen M, ed. (2007) d-side publications. 495–500. 15th European Symposium on Artficial Neural Networks (ESANN 2007).
[36] Simon HA. The architecture of complexity. Proceedings of the American Philosophical Society (1962) 106:467–482.
[37] Stanley KO, Miikkulainen R. Evolving neural networks through augmenting topologies. Evolutionary Computation (2002) 10(2):99–127.[CrossRef][Web of Science][Medline]
[38] Steil JJ. Online stability of backpropagation-decorrelation recurrent learning. Neurocomputing (2006) 69(7-9):642–650.[CrossRef][Web of Science]
[39] Steil JJ. Several ways to solve the MSO problem. Verleysen M, ed. (2007) Belgium: Evere. 489–494. In.
[40] Suttorp T, Hansen N, Igel C. Efficient covariance matrix update for variable metric evolution strategies. Machine Learning (2009) 75(2):167–197. doi: 10.1007/s10994-009-5102-1.[CrossRef][Web of Science]
[41] Torreele J. Temporal processing with recurrent networks: An evolutionary approach. Belew RK, Booker LB, eds. (1991) Morgan Kaufmann Publishers. 555–561. Proceedings of the Fourth International Conference on Genetic Algorithms (ICGA).
[42] Triesch J. Synergies between intrinsic and synaptic plasticity in individual model neurons. In: Advances in Neural Information Processing Systems 17—Saul LK, Weiss Y, Bottou L, eds. (2005) MIT Press. 1417–1424.
[43] Triesch J. Synergies between intrinsic and synaptic plasticity mechanisms. Neural Computation (2007) 19:885–909.[CrossRef][Web of Science][Medline]
[44] Vapnik VN. The Nature of Statistical Learning Theory (1995) Springer-Verlag.
[45] Verstraeten D, Schrauwen B, D’Haene M, Stroobandt D. An experimental unification of reservoir computing methods. Neural Networks (2007) 20(3):391–403.[CrossRef][Web of Science][Medline]
[46] Whitley D. The GENITOR algorithm and selection pressure: Why rank-based allocation of reproductive trials is best. (1989) 116–121. In.
[47] Wiegand S, Igel C, Handmann U. Evolutionary multi-objective optimization of neural networks for face detection. In: International Journal of Computational Intelligence and Applications (2004) 4:237–253. Special issue on Neurocomputing and Hybrid Methods for Evolving Intelligence.[CrossRef]
[48] Wieland AP. Evolving neural network controllers for unstable systems. (1991) 2:667–673. In International Joint Conference on Neural Networks (IJCNN 1991).
[49] Yao X. Evolving artificial neural networks. Proceedings of the IEEE (1999) 87(9):1423–1447.[CrossRef][Web of Science]
[50] Zitzler E, Thiele L. Multiobjective optimization using evolutionary algorithms – a comparative case study. Eiben A, Bäck T, Schoenauer M, Schwefel H-P, eds. (1998) Springer-Verlag. 292–301. In.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||