Abstract
In a reward-seeking task performed in a continuous environment, our previous work compared several Actor-Critic (AC) architectures implementing dopamine-like reinforcement learning mechanisms in the rat’s basal ganglia. The task complexity imposes the coordination of several AC submodules, each module being an expert trained in a particular subset of the task. We showed that the classical method where the choice of the expert to train at a given time depends on each expert’s performance suffered from strong limitations. We rather proposed to cluster the continuous state space by an ad hoc method that lacked autonomy and generalization abilities. In the present work we have combined the mixture of experts with self-organizing maps in order to cluster autonomously the experts’ responsibility space. On the one hand, we find that classical Kohonen maps give very variable results: some task decompositions provide very good and stable reinforcement learning performances, whereas some others are unadapted to the task. Moreover, they require the number of experts to be set a priori. On the other hand, algorithms like Growing Neural Gas or Growing When Required have the property to choose autonomously and incrementally the number of experts to train. They lead to good performances, even if they are still weaker than our hand-tuned task decomposition and than the best Kohonen maps that we got. We finally discuss on propositions about what information to add to these algorithms, such as knowledge of current behavior, in order to make the task decomposition appropriate to the reinforcement learning process.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Similar content being viewed by others
References
Albertin, S.V., Mulder, A.B., Tabuchi, E., Zugaro, M.B., Wiener, S.I.: Lesions of the medial shell of the nucleus accumbens impair rats in finding larger rewards, but spare reward-seeking behavior. Behavioral Brain Research 117(1-2), 173–183 (2000)
Arleo, A., Gerstner, W.: Spatial cognition and neuro-mimetic navigation: a model of hippo-campal place cell activity. Biological Cybernetics 83(3), 287–299 (2000)
Baldassarre, G.: A modular neural-network model of the basal ganglia’s role in learning and selecting motor behaviors. Journal of Cognitive Systems Research 3(1), 5–13 (2002)
Doya, K., Samejima., K., Katagiri, K., Kawato, M.: Multiple model-based reinforcement learning. Neural Computation 14(6), 1347–1369 (2002)
Filliat, D., Girard, B., Guillot, A., Khamassi, M., Lachèze, L., Meyer, J.-A.: State of the artificial rat Psikharpax. In: Schaal, S., Ijspeert, A., Billard, A., Vijayakumar, S., Hallam, J., Meyer, J.-A. (eds.) From Animals to Animats 8: Proceedings of the Seventh International Conference on Simulation of Adaptive Behavior, pp. 3–12. MIT Press, Cambridge (2004)
Fritzke, B.: A growing neural gas network learns topologies. In: Tesauro, G., Touretzkys, D.S., Leen, K. (eds.) Advances in Neural Information Processing Systems, pp. 625–632. MIT Press, Cambridge (1995)
Geman, S., Bienenstock, E., Doursat, R.: Neural networks and the bias/variance dilemma. Neural Computation 4, 1–58 (1992)
Gurney, K., Prescott, T.J., Redgrave, P.: A computational model of action selection in the basal ganglia. I. A new functional anatomy. Biological Cybernetics 84, 401–410 (2001)
Holmström, J.: Growing neural gas: Experiments with GNG, GNG with utility and supervised GNG. Master’s thesis, Uppsala University (2002)
Jog, M.S., Kubota, Y., Connolly, C.I., Hillegaart, V., Graybiel, A.M.: Building neural representations of habits. Science 286(5445), 1745–1749 (1999)
Khamassi, M., Lachèze, L., Girard, B., Berthoz, A., Guillot, A.: Actor-critic models of rein-forcement learning in the basal ganglia: From natural to artificial rats. Adaptive Behavior, Special Issue Towards Artificial Rodents 13(2), 131–148 (2005)
Kohonen, T.: Self-organizing maps. Springer, Heidelberg (1995)
Lee, J.K., Kim, I.H.: Reinforcement learning control using self-organizing map and multi-layer feed-forward neural network. In: International Conference on Control Automation and Systems, ICCAS 2003 (2003)
Meyer, J.-A., Guillot, A., Girard, B., Khamassi, M., Pirim, P., Berthoz, A.: The Psikharpax project: Towards building an artificial rat. Robotics and Autonomous Systems 50(4), 211–223 (2005)
Marsland, S., Shapiro, J., Nehmzow, U.: A self-organising network that grows when required. Neural Networks 15, 1041–1058 (2002)
Prescott, T.J., Redgrave, P., Gurney, K.: Layered control architectures in robots and vertebrates. Adaptive Behavior 7, 99–127 (1999)
Schultz, W., Dayan, P., Montague, P.R.: A neural substrate of prediction and reward. Science 275, 1593–1599 (1997)
Smith, A.J.: Applications of the self-organizing map to reinforcement learning. Neural Networks 15(8-9), 1107–1124 (2002)
Sutton, R.S., Barto, A.G.: Reinforcement learning: An introduction. The MIT Press, Cambridge (1998)
Tang, B., Heywood, M.I., Shepherd, M.: Input Partitioning to Mixture of Experts. In: IEEE/INNS International Joint Conference on Neural Networks, Honolulu, Hawaii, pp. 227–232 (2002)
Tani, J., Nolfi, S.: Learning to perceive the world as articulated: an approach for hierarchical learning in sensory-motor systems. Neural Networks 12, 1131–1141 (1999)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2006 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Khamassi, M., Martinet, LE., Guillot, A. (2006). Combining Self-organizing Maps with Mixtures of Experts: Application to an Actor-Critic Model of Reinforcement Learning in the Basal Ganglia. In: Nolfi, S., et al. From Animals to Animats 9. SAB 2006. Lecture Notes in Computer Science(), vol 4095. Springer, Berlin, Heidelberg. https://doi.org/10.1007/11840541_33
Download citation
DOI: https://doi.org/10.1007/11840541_33
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-38608-7
Online ISBN: 978-3-540-38615-5
eBook Packages: Computer ScienceComputer Science (R0)