Abstract
Primitives are essential features in human’s recognition capability. In this paper, we propose a method to form representations of motion primitives within a neural network model to learn sequences that consist of a combination of primitives. The training model is based on human development phenomena, namely motionese, where human infants learn simple motions first, and tend to learn complex motions as their skills develop. A neurodynamical model, Multiple Timescale Recurrent Neural Network (MTRNN), is used for the dynamics learning model. Neurons in MTRNN are composed hierarchically to learn different levels of information. The proposed method first trains the model using basic simple motions with neurons learning low level data. After training converges, neurons learning high level data are attached to the model, and complex motions are used to train the model. Experiments and analysis with drawing data show the models capability to form representations of primitives within the model.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Asada, M., MacDorman, K., Ishiguro, H., Kuniyoshi, Y.: Cognitive developmental robotics as a new paradigm for the design of humanoid robots. Rob. Auton. Syst. 37, 185–193 (2001)
Berk L.E., Winsler A.: Scaffolding Children’s Learning: Vygotsky and Early Childhood Education. National Association for Education (1995)
Triesch J., Teuscher C., Deák G.O., Carlson, E.: Gaze following: why (not) learn it? Dev. Sci. 9(2), 125–147 (2006)
Brand, R.J., Baldwin, D.A., Ashburn, L.A.: Evidence for ‘motionese’: modifications in motions’ infant-directed action. Dev. Sci. 5(1), 72–83 (2002)
Akgun, B., Cakmak, M., Yoo, J.W., Thoma, A.L.: Augmenting kinesthetic teaching with keyframes. In: Proceedings of International Conference on Machine Learning, Workshop on New Developments in Imitation Learning (2011)
Mohan, V., Morasso, P., Metta, G., Kasderidis, S.: Teaching a humanoid robot to draw ‘shapes’. Auton. Rob. 31(1), 21–53 (2011)
Luquet, G.H.: Le Dessin Enfantin (1927)
Mochizuki, K., Nishide, S., Okuno, H.G., Ogata, T.: Developmental human-robot imitation learning of drawing with a neuro dynamical system. In: Proceedings of IEEE International Conference on Systems, Man, and Cybernetics, pp. 2336–2341 (2013)
Nishide, S., Mochizuki, K., Okuno, H.G., Ogata, T.: Insertion of pause in drawing from babbling for robot’s developmental imitation learning. In: Proceedings of IEEE International Conference on Robotics and Automation, pp. 4785–4791 (2014)
Yamashita, Y., Tani, J.: Emergence of functional hierarchy in a multiple timescale recurrent neural network model: a humanoid Robot Experiment. PLoS Comput. Biol. 4(11), e1000220 (2008)
Acknowledgments
This research was supported by JSPS KAKENHI Grand Number 16H05877.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2020 Springer Nature Switzerland AG
About this paper
Cite this paper
Nomoto, F., Yasuda, T., Nishide, S., Kang, X., Ren, F. (2020). Hierarchical Learning of Primitives Using Neurodynamic Model. In: Fujita, H., Fournier-Viger, P., Ali, M., Sasaki, J. (eds) Trends in Artificial Intelligence Theory and Applications. Artificial Intelligence Practices. IEA/AIE 2020. Lecture Notes in Computer Science(), vol 12144. Springer, Cham. https://doi.org/10.1007/978-3-030-55789-8_62
Download citation
DOI: https://doi.org/10.1007/978-3-030-55789-8_62
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-55788-1
Online ISBN: 978-3-030-55789-8
eBook Packages: Computer ScienceComputer Science (R0)