Abstract
This paper shows that active-inference-based, flexible, adaptive goal-directed behavior can be generated by utilizing temporal gradients in a recurrent neural network (RNN). The RNN learns a dynamical sensorimotor forward model of a partially observable environment. It then uses this model to execute goal-directed policy inference online. The internal neural activities encode the predictive state of the controlled entity. The active inference process projects these activities into the future via the RNN’s recurrences, following a tentative sequence of motor commands. This sequence is adapted by back-projecting error between the forward-projected hypothetical states and the desired goal states onto the motor commands. As an example, we show that a trained RNN model can be used to precisely control a multi-copter-like system. Moreover, we show that the RNN can plan hundreds of time steps ahead, unfolding non-linear imaginary paths around obstacles.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Butz, M.V., Herbort, O., Hoffmann, J.: Exploiting redundancy for flexible behavior: unsupervised learning in a modular sensorimotor control architecture. Psychol. Rev. 114, 1015–1046 (2007)
Butz, M.V.: Towards a unified sub-symbolic computational theory of cognition. Fronti. Psychol. 7(925) (2016)
Friston, K.: The free-energy principle: a rough guide to the brain? Trends Cogn. Sci. 13(7), 293–301 (2009)
Friston, K., FitzGerald, T., Rigoli, F., Schwartenbeck, P., Pezzulo, G.: Active inference: a process theory. Neural Comput. 29(1), 1–49 (2016)
Friston, K.J., Daunizeau, J., Kilner, J., Kiebel, S.J.: Action and behavior: a free-energy formulation. Biol. Cybern. 102(3), 227–260 (2010)
Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 9(8), 1735–1780 (1997)
Jordan, M.I., Rumelhart, D.E.: Forward models: supervised learning with a distal teacher. Cogn. Sci. 16, 307–354 (1992)
Kingma, D.P., Ba, J.L.: Adam: a method for stochastic optimization. In: 3rd International Conference for Learning Representations abs/1412.6980 (2015)
Otte, S., Krechel, D., Liwicki, M.: JANNLab neural network framework for Java. In: Poster Proceedings MLDM 2013, pp. 39–46. ibai-publishing, New York (2013)
Otte, S., Liwicki, M., Zell, A.: Dynamic cortex memory: enhancing recurrent neural networks for gradient-based sequence learning. In: Wermter, S., Weber, C., Duch, W., Honkela, T., Koprinkova-Hristova, P., Magg, S., Palm, G., Villa, A.E.P. (eds.) ICANN 2014. LNCS, vol. 8681, pp. 1–8. Springer, Cham (2014). doi:10.1007/978-3-319-11179-7_1
Otte, S., Liwicki, M., Zell, A.: An analysis of dynamic cortex memory networks. In: International Joint Conference on Neural Networks (IJCNN), pp. 3338–3345. Killarney, Ireland, Jul 2015
Otte, S., Zwiener, A., Hanten, R., Zell, A.: Inverse recurrent models – an application scenario for many-joint robot arm control. In: Villa, A.E.P., Masulli, P., Pons Rivero, A.J. (eds.) ICANN 2016. LNCS, vol. 9886, pp. 149–157. Springer, Cham (2016). doi:10.1007/978-3-319-44778-0_18
Sutton, R.S., Barto, A.G.: Reinforcement Learning: An Introduction (1998)
Werbos, P.: Backpropagation through time: what it does and how to do it. Proc. IEEE 78(10), 1550–1560 (1990)
Wolpert, D.M., Kawato, M.: Multiple paired forward and inverse models for motor control. Neural Netw. 11, 1317–1329 (1998)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2017 Springer International Publishing AG
About this paper
Cite this paper
Otte, S., Schmitt, T., Friston, K., Butz, M.V. (2017). Inferring Adaptive Goal-Directed Behavior Within Recurrent Neural Networks. In: Lintas, A., Rovetta, S., Verschure, P., Villa, A. (eds) Artificial Neural Networks and Machine Learning – ICANN 2017. ICANN 2017. Lecture Notes in Computer Science(), vol 10613. Springer, Cham. https://doi.org/10.1007/978-3-319-68600-4_27
Download citation
DOI: https://doi.org/10.1007/978-3-319-68600-4_27
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-68599-1
Online ISBN: 978-3-319-68600-4
eBook Packages: Computer ScienceComputer Science (R0)