Development of proactive and reactive behavior via meta-learning of prediction error variance

Shingo Murata; Jun Namikawa; Hiroaki Arie; Jun Tani; Shigeki Sugano

doi:10.1007/978-3-642-42054-2_67

Development of proactive and reactive behavior via meta-learning of prediction error variance

Shingo Murata, Jun Namikawa, Hiroaki Arie, Jun Tani, Shigeki Sugano

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

This paper investigates a possible neurodynamic mechanism that enables autonomous switching between two basic behavioral modes, namely a "proactive mode" and a "reactive mode." In the proactive mode, actions are generated as intended, whereas in the reactive mode actions are generated in response to the sensory state.We conducted neurorobotics experiments to investigate how these two modes can develop and how a robot can learn to switch autonomously between the two modes as necessary by utilizing our recently developed dynamic neural network model. Tasks designed for the robot included switching between proactive imitation of other's predictable movements using acquired memories and reactive following of other's unpredictable movements through iterative learning of alternating predictable and unpredictable movement patterns. The experimental results revealed that this "meta-learning" capability can lead to self-organization of adequate contextual dynamical structures that can perform autonomous switching between the different behavioral modes.

Original language	English
Title of host publication	Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings
Pages	537-544
Number of pages	8
Edition	PART 1
DOIs	https://doi.org/10.1007/978-3-642-42054-2_67
Publication status	Published - 2013 Dec 1
Externally published	Yes
Event	20th International Conference on Neural Information Processing, ICONIP 2013 - Daegu, Korea, Republic of Duration: 2013 Nov 3 → 2013 Nov 7

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 1
Volume	8226 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	20th International Conference on Neural Information Processing, ICONIP 2013
Country/Territory	Korea, Republic of
City	Daegu
Period	13/11/3 → 13/11/7

Keywords

Humanoid robot
Neurorobotics
Recurrent neural network

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-642-42054-2_67

Cite this

Murata, S., Namikawa, J., Arie, H., Tani, J., & Sugano, S. (2013). Development of proactive and reactive behavior via meta-learning of prediction error variance. In Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings (PART 1 ed., pp. 537-544). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8226 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-42054-2_67

Development of proactive and reactive behavior via meta-learning of prediction error variance. / Murata, Shingo; Namikawa, Jun; Arie, Hiroaki et al.
Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings. PART 1. ed. 2013. p. 537-544 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8226 LNCS, No. PART 1).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Murata, S, Namikawa, J, Arie, H, Tani, J & Sugano, S 2013, Development of proactive and reactive behavior via meta-learning of prediction error variance. in Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings. PART 1 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 8226 LNCS, pp. 537-544, 20th International Conference on Neural Information Processing, ICONIP 2013, Daegu, Korea, Republic of, 13/11/3. https://doi.org/10.1007/978-3-642-42054-2_67

Murata S, Namikawa J, Arie H, Tani J, Sugano S. Development of proactive and reactive behavior via meta-learning of prediction error variance. In Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings. PART 1 ed. 2013. p. 537-544. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). doi: 10.1007/978-3-642-42054-2_67

Murata, Shingo ; Namikawa, Jun ; Arie, Hiroaki et al. / Development of proactive and reactive behavior via meta-learning of prediction error variance. Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings. PART 1. ed. 2013. pp. 537-544 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).

@inproceedings{75eefbdb4ca046f7935f6103a5903d78,

title = "Development of proactive and reactive behavior via meta-learning of prediction error variance",

abstract = "This paper investigates a possible neurodynamic mechanism that enables autonomous switching between two basic behavioral modes, namely a {"}proactive mode{"} and a {"}reactive mode.{"} In the proactive mode, actions are generated as intended, whereas in the reactive mode actions are generated in response to the sensory state.We conducted neurorobotics experiments to investigate how these two modes can develop and how a robot can learn to switch autonomously between the two modes as necessary by utilizing our recently developed dynamic neural network model. Tasks designed for the robot included switching between proactive imitation of other's predictable movements using acquired memories and reactive following of other's unpredictable movements through iterative learning of alternating predictable and unpredictable movement patterns. The experimental results revealed that this {"}meta-learning{"} capability can lead to self-organization of adequate contextual dynamical structures that can perform autonomous switching between the different behavioral modes.",

keywords = "Humanoid robot, Neurorobotics, Recurrent neural network",

author = "Shingo Murata and Jun Namikawa and Hiroaki Arie and Jun Tani and Shigeki Sugano",

year = "2013",

month = dec,

day = "1",

doi = "10.1007/978-3-642-42054-2_67",

language = "English",

isbn = "9783642420535",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

number = "PART 1",

pages = "537--544",

booktitle = "Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings",

edition = "PART 1",

note = "20th International Conference on Neural Information Processing, ICONIP 2013 ; Conference date: 03-11-2013 Through 07-11-2013",

}

TY - GEN

T1 - Development of proactive and reactive behavior via meta-learning of prediction error variance

AU - Murata, Shingo

AU - Namikawa, Jun

AU - Arie, Hiroaki

AU - Tani, Jun

AU - Sugano, Shigeki

PY - 2013/12/1

Y1 - 2013/12/1

N2 - This paper investigates a possible neurodynamic mechanism that enables autonomous switching between two basic behavioral modes, namely a "proactive mode" and a "reactive mode." In the proactive mode, actions are generated as intended, whereas in the reactive mode actions are generated in response to the sensory state.We conducted neurorobotics experiments to investigate how these two modes can develop and how a robot can learn to switch autonomously between the two modes as necessary by utilizing our recently developed dynamic neural network model. Tasks designed for the robot included switching between proactive imitation of other's predictable movements using acquired memories and reactive following of other's unpredictable movements through iterative learning of alternating predictable and unpredictable movement patterns. The experimental results revealed that this "meta-learning" capability can lead to self-organization of adequate contextual dynamical structures that can perform autonomous switching between the different behavioral modes.

AB - This paper investigates a possible neurodynamic mechanism that enables autonomous switching between two basic behavioral modes, namely a "proactive mode" and a "reactive mode." In the proactive mode, actions are generated as intended, whereas in the reactive mode actions are generated in response to the sensory state.We conducted neurorobotics experiments to investigate how these two modes can develop and how a robot can learn to switch autonomously between the two modes as necessary by utilizing our recently developed dynamic neural network model. Tasks designed for the robot included switching between proactive imitation of other's predictable movements using acquired memories and reactive following of other's unpredictable movements through iterative learning of alternating predictable and unpredictable movement patterns. The experimental results revealed that this "meta-learning" capability can lead to self-organization of adequate contextual dynamical structures that can perform autonomous switching between the different behavioral modes.

KW - Humanoid robot

KW - Neurorobotics

KW - Recurrent neural network

UR - http://www.scopus.com/inward/record.url?scp=84893353213&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84893353213&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-42054-2_67

DO - 10.1007/978-3-642-42054-2_67

M3 - Conference contribution

AN - SCOPUS:84893353213

SN - 9783642420535

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 537

EP - 544

BT - Neural Information Processing - 20th International Conference, ICONIP 2013, Proceedings

T2 - 20th International Conference on Neural Information Processing, ICONIP 2013

Y2 - 3 November 2013 through 7 November 2013

ER -

Development of proactive and reactive behavior via meta-learning of prediction error variance

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this