Advertisement

Cumulative Learning Through Intrinsic Reinforcements

  • Vieri G. SantucciEmail author
  • Gianluca Baldassarre
  • Marco Mirolli

Abstract

Building artificial agents able to autonomously learn new skills and to easily adapt in different and complex environments is an important goal for robotics and machine learning. We propose that providing reinforcement learning artificial agents with a learning signal that resembles the characteristic of the phasic activations of dopaminergic neurons would be an advancement in the development of more autonomous and versatile systems. In particular, we suggest that the particular composition of such a signal, determined by both extrinsic and intrinsic reinforcements, would be suitable to improve the implementation of cumulative learning in artificial agents. To validate our hypothesis we performed experiments with a simulated robotic system that has to learn different skills to obtain extrinsic rewards. We compare different versions of the system varying the composition of the learning signal and we show that the only system able to reach high performance in the task is the one that implements the learning signal suggested by our hypothesis.

Keywords

Reinforcement Learning Artificial Agent Output Unit Input Unit Learning Signal 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Notes

Acknowledgements

This research was supported by the European Community 7th Framework Programme (FP7/2007-2013), “Challenge 2 - Cognitive Systems, Interaction, Robotics”, grant agreement No. ICT-IP-231722, project “IM-CLeVeR - Intrinsically Motivated Cumulative Learning Versatile Robots”.

References

  1. 1.
    Weng, J., McClelland, J., Pentland, A., Sporns, O., Stockman, I., Sur, M., Thelen, E.: Artificial intelligence. Autonomous mental development by robots and animals. Science 291(5504), 599–600 (2001)Google Scholar
  2. 2.
    Baldassarre, G., Mirolli, M.: What are the key open challenges for understanding autonomous cumulative learning of skills? Auton. Mental Develop. Newsl. 7(2), 2–9 (2010)Google Scholar
  3. 3.
    White, R.: Motivation reconsidered: The concept of competence. Psychol. Rev. 66, 297–333 (1959)CrossRefGoogle Scholar
  4. 4.
    Berlyne, D.: Conflict, Arousal and Curiosity. McGraw Hill, New York (1960)Google Scholar
  5. 5.
    Ryan, R.M., Deci, E.L.: Intrinsic and extrinsic motivations: Classic definitions and new directions. Contemp. Educ. Psychol. 25(1), 54–67 (2000)CrossRefGoogle Scholar
  6. 6.
    Montgomery, K.: The role of the exploratory drive in learning. J. Comp. Psychol. 47(1), 60–64 (1954)CrossRefGoogle Scholar
  7. 7.
    Butler, R.A., Harlow, H.F.: Discrimination learning and learning sets to visual exploration incentives. J. Gen. Psychol. 57(2), 257–264 (1957)CrossRefGoogle Scholar
  8. 8.
    Hull, C.L.: Principles of behavior. Appleton-century-crofts, New York (1943)Google Scholar
  9. 9.
    Kish, G.B.: Learning when the onset of illumination is used as reinforcing stimulus. J. Comp. Physiol. Psychol. 48(4), 261–264 (1955)CrossRefGoogle Scholar
  10. 10.
    Glow, P., Winefield, A.: Response-contingent sensory change in a causally structured environment. Learn. Behav. 6, 1–18 (1978)CrossRefGoogle Scholar
  11. 11.
    Reed, P., Mitchell, C., Nokes, T.: Intrinsic reinforcing properties of putatively neutral stimuli in an instrumental two-lever discrimination task. Anim. Lear. Behav. 24, 38–45 (1996)CrossRefGoogle Scholar
  12. 12.
    Schmidhuber, J.: Curious model-building control system. In: Proceedings of International Joint Conference on Neural Networks, vol. 2, pp. 1458–1463. IEEE, Singapore (1991)Google Scholar
  13. 13.
    Huang, X., Weng, J.: Novelty and reinforcement learning in the value system of developmental robots. In: Prince, C., Demiris, Y., Marom, Y., Kozima, H., Balkenius, C. (eds.) Proceedings of the Second International Workshop on Epigenetic Robotics, vol. 94, pp. 47–55. Lund University (2002)Google Scholar
  14. 14.
    Oudeyer, P., Kaplan, F., Hafner, V.: Intrinsic motivation system for autonomous mental development. IEEE T. Evolut. Comput. 11, 703–713 (2007)CrossRefGoogle Scholar
  15. 15.
    Baranes, A., Oudeyer, P.Y.: Intrinsically motivated goal exloration for active motor learning in robots: A case study. In: Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, Taipei, Taiwan (2010)Google Scholar
  16. 16.
    Barto, A., Singh, S., Chantanez, N.: Intrinsically motivated learning of hierarchical collections of skills. In: Proceedings of the Third International Conference on Developmental Learning (ICDL), pp. 112–119 (2004)Google Scholar
  17. 17.
    Schembri, M., Mirolli, M., Baldassarre, G.: Evolving internal reinforcers for an intrinsically motivated reinforcement-learning robot. In: Demiris, Y., Mareschal, D., Scassellati, B., Weng, J. (eds.) Proceedings of the 6th International Conference on Development and Learning, pp. E1–6. Imperial College, London (2007)Google Scholar
  18. 18.
    Oudeyer, P.Y., Kaplan, F.: What is intrinsic motivation? A typology of computational approaches. Front. Neurorobot. 1(1), 1–14 (2007)Google Scholar
  19. 19.
    Baldassarre, G., Mirolli, M. (eds.): Intrinsically Motivated Learning in Natural and Artificial Systems. Springer, Berlin (2013)Google Scholar
  20. 20.
    Santucci, V.G., Baldassarre, G., Mirolli, M.: Biological cumulative learning through intrinsic motivation: A simulated robotic study on the development of visually-guided reaching. In: Johansson, B., Sahin, E., Balkenius, C. (eds.) Proceedings of the Tenth International Conference on Epigenetic Robotics, pp. 121–128. Lund University Cognitive Studies, Lund (2010)Google Scholar
  21. 21.
    Wise, R.: Dopamine, learning and motivation. Nat. Rev. Neurosci. 5(6), 483–494 (2004)CrossRefMathSciNetGoogle Scholar
  22. 22.
    Schultz, W.: Behavioral theories and the neurophysiology of reward. Annu. Rev. Psychol. 57, 87–115 (2006)CrossRefGoogle Scholar
  23. 23.
    Berridge, K.: The debate over dopamine’s role in reward: The case for incentive salience. Psychopharmacology. 191(3), 391–431 (2007)CrossRefGoogle Scholar
  24. 24.
    Romo, R., Schultz, W.: Dopamine neurons of the monkey midbrain: Contingencies of responses to active touch during self-initiated arm movements. J. Neurophysiol. 63(3), 592–606 (1990)Google Scholar
  25. 25.
    Ljungberg, T., Apicella, P., Schultz, W.: Responses of monkey midbrain dopamine neurons during delayed alternation performance. Brain Res. 567(2), 337–341 (1991)CrossRefGoogle Scholar
  26. 26.
    Schultz, W., Apicella, P., Ljumberg, T.: Responses of monkey dopamine neurons to reaward and conditioned stimuli during successive steps of learning a delayed response task. J. Neurosci. 13, 900–913 (1993)Google Scholar
  27. 27.
    Mirenowicz, J., Schultz, W.: Importance of unpredictability for reward responses in primate dopamine neurons. J. Neurophysiol. 72(2), 1024–1027 (1994)Google Scholar
  28. 28.
    Ljungberg, T., Apicella, P., Schultz, W.: Responses of monkey dopamine neurons during learning of behavioral reactions. J. Neurophysiol. 67(1), 145–163 (1992)Google Scholar
  29. 29.
    Schultz, W.: Predictive reward signal of dopamine neurons. J. Neurophysiol. 80(1), 1–27 (1998)Google Scholar
  30. 30.
    Horvitz, J.C.: Mesolimbocortical and nigrostriatal dopamine responses to salient non-reward events. Neuroscience 96(4), 651–656 (2000)CrossRefGoogle Scholar
  31. 31.
    Dommett, E., Coizet, V., Blaha, C.D., Martindale, J., Lefebvre, V., Walton, N., Mayhew, J.E.W., Overton, P.G., Redgrave, P.: How visual stimuli activate dopaminergic neurons at short latency. Science 307(5714), 1476–1479 (2005)CrossRefGoogle Scholar
  32. 32.
    Houk, J., Adams, J., Barto, A.: A model of how the basal ganglia generate and use neural signals that predict reinforcement. In: Houk, J.C., Davis, J.L., Beiser, D.G. (eds.) Models of Information Processing in the Basal Ganglia, pp. 249–270. MIT Press, Cambridge (1995)Google Scholar
  33. 33.
    Schultz, W., Dayan, P., Montague, P.R.: A neural substrate of prediction and reward. Science 275(5306), 1593–1599 (1997)CrossRefGoogle Scholar
  34. 34.
    Sutton, R., Barto, A.: Reinforcement Learning: An Introduction. MIT Press, Cambridge (1998)Google Scholar
  35. 35.
    Redgrave, P., Gurney, K.: The short-latency dopamine signal: A role in discovering novel actions? Nat. Rev. Neurosci. 7(12), 967–975 (2006)CrossRefGoogle Scholar
  36. 36.
    Redgrave, P., Vautrelle, N., Reynolds, J.N.J.: Functional properties of the basal ganglia’s re-entrant loop architecture: Selection and reinforcement. Neuroscience. 198, 138–151 (2011)CrossRefGoogle Scholar
  37. 37.
    Mirolli, M., Santucci, V.G., Baldassarre, G.: Phasic dopamine as a prediction error of intrinsic and extrinsic reinforcements driving both action acquisition and reward maximization: a simulated robotic study. Neural Networks 39, 40–51 (2013)CrossRefGoogle Scholar
  38. 38.
    Baldassarre, G.: What are intrinsic motivations? A biological perspective. In: Cangelosi, A., Triesch, J., Fasel, I., Rohlfing, K., Nori, F., Oudeyer, P.Y., Schlesinger, M., Nagai, Y. (eds.) Proceedings of the International Conference on Development and Learning and Epigenetic Robotics (ICDL-EpiRob-2011), pp. E1–8. IEEE, Piscataway (2011)CrossRefGoogle Scholar
  39. 39.
    Romanelli, P., Esposito, V., Schaal, D.W., Heit, G.: Somatotopy in the basal ganglia: Experimental and clinical evidence for segregated sensorimotor channels. Brain Res. Rev. 48(1), 112–128 (2005)CrossRefGoogle Scholar
  40. 40.
    Graybiel, A.M.: The basal ganglia: Learning new tricks and loving it. Curr. Opin. Neurobiol. 15(6), 638–644 (2005)CrossRefGoogle Scholar
  41. 41.
    Joel, D., Niv, Y., Ruppin, E.: Actor-critic models of the basal ganglia: New anatomical and computational perspectives. Neural Network 15(4–6), 535–547 (2002)CrossRefGoogle Scholar
  42. 42.
    Sutton, R.S., McAllester, D.A., Singh, S.P., Mansour, Y.: Policy gradient methods for reinforcement learning with function approximation. In: NIPS, vol. 99, pp. 1057–1063 (1999)Google Scholar
  43. 43.
    Schultz, W.: Getting formal with dopamine and reward. Neuron 36(2), 241–263 (2002)CrossRefGoogle Scholar
  44. 44.
    Pouget, A., Snyder, L.H.: Computational approaches to sensorimotor transformations. Nat. Neurosci. 3(Suppl), 1192–1198 (2000)CrossRefGoogle Scholar
  45. 45.
    Buhmann, M.: Radial Basis Functions. Cambridge University Press, New York (2003)CrossRefzbMATHGoogle Scholar
  46. 46.
    Sutton, R., Tanner, B.: Temporal-difference networks. Adv. Neural Inf. Process. Syst. 17, 1377–1348 (2005)Google Scholar
  47. 47.
    Singh, S., Lewis, R., Barto, A., Sorg, J.: Intrinsically motivated reinforcement learning: An evolutionary perspective. IEEE T. Auton. Mental Dev. 2(2), 70–82 (2010)CrossRefGoogle Scholar
  48. 48.
    Doya, K., Samejima, K., Katagiri, K., Kawato, M.: Multiple model-based reinforcement learning. Neural Comput. 14(6), 1347–1369 (2002)CrossRefzbMATHGoogle Scholar
  49. 49.
    Caligiore, D., Mirolli, M., Parisi, D., Baldassarre, G.: A bioinspired hierarchical reinforcement learning architecture for modeling learning of multiple skills with continuous states and actions. In: Johansson, B., Sahin, E., Balkenius, C. (eds.) Proceedings of the Tenth International Conference on Epigenetic Robotics, pp. 27–34 (2010)Google Scholar
  50. 50.
    Schembri, M., Mirolli, M., Baldassarre, G.: Evolving childhood’s length and learning parameters in an intrinsically motivated reinforcement learning robot. In: Berthouze, L., Dhristiopher, G., Littman, M., Kozima, H., Balkenius, C. (eds.) Proceedings of the Seventh International Conference on Epigenetic Robotics, pp. 141–148. Lund University Cognitive Studies, Lund (2007)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Vieri G. Santucci
    • 1
    • 2
    Email author
  • Gianluca Baldassarre
    • 1
  • Marco Mirolli
    • 1
  1. 1.Istituto di Scienze e Tecnologie della Cognizione (ISTC), Consiglio Nazionale delle Ricerche (CNR)Laboratory of Computational Embodied Neuroscience (LOCEN)RomaItalia
  2. 2.School of Computing and MathematicsUniversity of PlymouthPlymouthUK

Personalised recommendations