Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines

H. Bou Ammar; D.C. Mocanu; M.E. Taylor; K. Driessens; K.P. Tuyls; G. Weiss

doi:10.1007/978-3-642-40991-2_29

Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines

H. Bou Ammar, D.C. Mocanu, M.E. Taylor, K. Driessens, K.P. Tuyls, G. Weiss

Electro-Optical Communication

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

17 Citations (Scopus)

91 Downloads (Pure)

Abstract

Existing reinforcement learning approaches are often hampered by learning tabula rasa. Transfer for reinforcement learning tackles this problem by enabling the reuse of previously learned results, but may require an inter-task mapping to encode how the previously learned task and the new task are related. This paper presents an autonomous framework for learning inter-task mappings based on an adaptation of restricted Boltzmann machines. Both a full model and a computationally efficient factored model are introduced and shown to be effective in multiple transfer learning scenarios.

Original language	English
Title of host publication	Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II
Editors	H. Blockeel, K. Kersting, S. Nijssen, F. Zelezny
Place of Publication	Berlin
Publisher	Springer
Pages	449-464
ISBN (Print)	978-3-642-40990-5
DOIs	https://doi.org/10.1007/978-3-642-40991-2_29
Publication status	Published - 2013
Event	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, September 23-27, 2013, Prague, Czech Republic - Prague, Czech Republic Duration: 23 Sep 2013 → 27 Sep 2013 http://www.ecmlpkdd2013.org/

Publication series

Name	Lecture Notes in Computer Science
Volume	8189
ISSN (Print)	0302-9743

Conference

Conference	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, September 23-27, 2013, Prague, Czech Republic
Abbreviated title	ECMLPKDD 2013
Country/Territory	Czech Republic
City	Prague
Period	23/09/13 → 27/09/13
Internet address	http://www.ecmlpkdd2013.org/

Access to Document

10.1007/978-3-642-40991-2_29

396383893296400Final author version , 527 KB

Cite this

Bou Ammar, H., Mocanu, D. C., Taylor, M. E., Driessens, K., Tuyls, K. P., & Weiss, G. (2013). Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines. In H. Blockeel, K. Kersting, S. Nijssen, & F. Zelezny (Eds.), Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II (pp. 449-464). (Lecture Notes in Computer Science; Vol. 8189). Springer. https://doi.org/10.1007/978-3-642-40991-2_29

Bou Ammar, H. ; Mocanu, D.C. ; Taylor, M.E. ; Driessens, K. ; Tuyls, K.P. ; Weiss, G. / Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines. Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II. editor / H. Blockeel ; K. Kersting ; S. Nijssen ; F. Zelezny. Berlin : Springer, 2013. pp. 449-464 (Lecture Notes in Computer Science).

@inproceedings{f3e146a6c72b4026a966a5ad84dec67b,

title = "Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines",

abstract = "Existing reinforcement learning approaches are often hampered by learning tabula rasa. Transfer for reinforcement learning tackles this problem by enabling the reuse of previously learned results, but may require an inter-task mapping to encode how the previously learned task and the new task are related. This paper presents an autonomous framework for learning inter-task mappings based on an adaptation of restricted Boltzmann machines. Both a full model and a computationally efficient factored model are introduced and shown to be effective in multiple transfer learning scenarios.",

author = "{Bou Ammar}, H. and D.C. Mocanu and M.E. Taylor and K. Driessens and K.P. Tuyls and G. Weiss",

year = "2013",

doi = "10.1007/978-3-642-40991-2_29",

language = "English",

isbn = "978-3-642-40990-5",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "449--464",

editor = "H. Blockeel and K. Kersting and S. Nijssen and F. Zelezny",

booktitle = "Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II",

address = "Germany",

note = "European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, September 23-27, 2013, Prague, Czech Republic, ECMLPKDD 2013 ; Conference date: 23-09-2013 Through 27-09-2013",

url = "http://www.ecmlpkdd2013.org/",

}

Bou Ammar, H, Mocanu, DC, Taylor, ME, Driessens, K, Tuyls, KP & Weiss, G 2013, Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines. in H Blockeel, K Kersting, S Nijssen & F Zelezny (eds), Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II. Lecture Notes in Computer Science, vol. 8189, Springer, Berlin, pp. 449-464, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, September 23-27, 2013, Prague, Czech Republic, Prague, Czech Republic, 23/09/13. https://doi.org/10.1007/978-3-642-40991-2_29

Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines. / Bou Ammar, H.; Mocanu, D.C.; Taylor, M.E.; Driessens, K.; Tuyls, K.P.; Weiss, G.

Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II. ed. / H. Blockeel; K. Kersting; S. Nijssen; F. Zelezny. Berlin : Springer, 2013. p. 449-464 (Lecture Notes in Computer Science; Vol. 8189).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines

AU - Bou Ammar, H.

AU - Mocanu, D.C.

AU - Taylor, M.E.

AU - Driessens, K.

AU - Tuyls, K.P.

AU - Weiss, G.

PY - 2013

Y1 - 2013

N2 - Existing reinforcement learning approaches are often hampered by learning tabula rasa. Transfer for reinforcement learning tackles this problem by enabling the reuse of previously learned results, but may require an inter-task mapping to encode how the previously learned task and the new task are related. This paper presents an autonomous framework for learning inter-task mappings based on an adaptation of restricted Boltzmann machines. Both a full model and a computationally efficient factored model are introduced and shown to be effective in multiple transfer learning scenarios.

AB - Existing reinforcement learning approaches are often hampered by learning tabula rasa. Transfer for reinforcement learning tackles this problem by enabling the reuse of previously learned results, but may require an inter-task mapping to encode how the previously learned task and the new task are related. This paper presents an autonomous framework for learning inter-task mappings based on an adaptation of restricted Boltzmann machines. Both a full model and a computationally efficient factored model are introduced and shown to be effective in multiple transfer learning scenarios.

U2 - 10.1007/978-3-642-40991-2_29

DO - 10.1007/978-3-642-40991-2_29

M3 - Conference contribution

SN - 978-3-642-40990-5

T3 - Lecture Notes in Computer Science

SP - 449

EP - 464

BT - Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II

A2 - Blockeel, H.

A2 - Kersting, K.

A2 - Nijssen, S.

A2 - Zelezny, F.

PB - Springer

CY - Berlin

T2 - European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, September 23-27, 2013, Prague, Czech Republic

Y2 - 23 September 2013 through 27 September 2013

ER -

Bou Ammar H, Mocanu DC, Taylor ME, Driessens K, Tuyls KP, Weiss G. Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines. In Blockeel H, Kersting K, Nijssen S, Zelezny F, editors, Machine Learning and Knowledge Discovery in Databases : European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part II. Berlin: Springer. 2013. p. 449-464. (Lecture Notes in Computer Science). https://doi.org/10.1007/978-3-642-40991-2_29

Automatically mapped transfer between reinforcement learning tasks via three-way restricted Boltzmann machines

Abstract

Publication series

Conference

Access to Document

Fingerprint

Cite this