Learning features and predictive transformation encoding based on a horizontal product model

Junpei Zhong, Cornelius Weber, Stefan Wermter

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Citations (Scopus)

Abstract

The visual system processes the features and movement of an object in separate pathways, called the ventral and dorsal streams. To integrate this principle in a functional model, a recurrent predictive network with a horizontal product is introduced. Learned in an unsupervised manner, two sets of hidden units represent cells in the ventral and dorsal pathways, respectively. Experiments show that the activity in the ventral-like units persists, given that the same feature appears in the receptive field, whilst the activity in the dorsal-like units shows a fluctuating pattern with different directions of object movements. Moreover, we show that the position information predicts the input's future position taking into account its moving direction due to the direction-selective responses of the dorsal-like units.

Original languageEnglish
Title of host publicationArtificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings
Pages539-546
Number of pages8
Volume7552 LNCS
EditionPART 1
DOIs
Publication statusPublished - 2012
Externally publishedYes
Event22nd International Conference on Artificial Neural Networks, ICANN 2012 - Lausanne, Switzerland
Duration: 2012 Sep 112012 Sep 14

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume7552 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other22nd International Conference on Artificial Neural Networks, ICANN 2012
CountrySwitzerland
CityLausanne
Period12/9/1112/9/14

Fingerprint

Product Model
Encoding
Horizontal
Unit
Pathway
Experiments
Receptive Field
Visual System
Functional Model
Integrate
Predict
Learning
Cell
Experiment
Object
Movement

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Zhong, J., Weber, C., & Wermter, S. (2012). Learning features and predictive transformation encoding based on a horizontal product model. In Artificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings (PART 1 ed., Vol. 7552 LNCS, pp. 539-546). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7552 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-33269-2_68

Learning features and predictive transformation encoding based on a horizontal product model. / Zhong, Junpei; Weber, Cornelius; Wermter, Stefan.

Artificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings. Vol. 7552 LNCS PART 1. ed. 2012. p. 539-546 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7552 LNCS, No. PART 1).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zhong, J, Weber, C & Wermter, S 2012, Learning features and predictive transformation encoding based on a horizontal product model. in Artificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings. PART 1 edn, vol. 7552 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 7552 LNCS, pp. 539-546, 22nd International Conference on Artificial Neural Networks, ICANN 2012, Lausanne, Switzerland, 12/9/11. https://doi.org/10.1007/978-3-642-33269-2_68
Zhong J, Weber C, Wermter S. Learning features and predictive transformation encoding based on a horizontal product model. In Artificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings. PART 1 ed. Vol. 7552 LNCS. 2012. p. 539-546. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-642-33269-2_68
Zhong, Junpei ; Weber, Cornelius ; Wermter, Stefan. / Learning features and predictive transformation encoding based on a horizontal product model. Artificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings. Vol. 7552 LNCS PART 1. ed. 2012. pp. 539-546 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).
@inproceedings{a5e2e0e798e548a6b93fc8e401fc8bca,
title = "Learning features and predictive transformation encoding based on a horizontal product model",
abstract = "The visual system processes the features and movement of an object in separate pathways, called the ventral and dorsal streams. To integrate this principle in a functional model, a recurrent predictive network with a horizontal product is introduced. Learned in an unsupervised manner, two sets of hidden units represent cells in the ventral and dorsal pathways, respectively. Experiments show that the activity in the ventral-like units persists, given that the same feature appears in the receptive field, whilst the activity in the dorsal-like units shows a fluctuating pattern with different directions of object movements. Moreover, we show that the position information predicts the input's future position taking into account its moving direction due to the direction-selective responses of the dorsal-like units.",
author = "Junpei Zhong and Cornelius Weber and Stefan Wermter",
year = "2012",
doi = "10.1007/978-3-642-33269-2_68",
language = "English",
isbn = "9783642332685",
volume = "7552 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 1",
pages = "539--546",
booktitle = "Artificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings",
edition = "PART 1",

}

TY - GEN

T1 - Learning features and predictive transformation encoding based on a horizontal product model

AU - Zhong, Junpei

AU - Weber, Cornelius

AU - Wermter, Stefan

PY - 2012

Y1 - 2012

N2 - The visual system processes the features and movement of an object in separate pathways, called the ventral and dorsal streams. To integrate this principle in a functional model, a recurrent predictive network with a horizontal product is introduced. Learned in an unsupervised manner, two sets of hidden units represent cells in the ventral and dorsal pathways, respectively. Experiments show that the activity in the ventral-like units persists, given that the same feature appears in the receptive field, whilst the activity in the dorsal-like units shows a fluctuating pattern with different directions of object movements. Moreover, we show that the position information predicts the input's future position taking into account its moving direction due to the direction-selective responses of the dorsal-like units.

AB - The visual system processes the features and movement of an object in separate pathways, called the ventral and dorsal streams. To integrate this principle in a functional model, a recurrent predictive network with a horizontal product is introduced. Learned in an unsupervised manner, two sets of hidden units represent cells in the ventral and dorsal pathways, respectively. Experiments show that the activity in the ventral-like units persists, given that the same feature appears in the receptive field, whilst the activity in the dorsal-like units shows a fluctuating pattern with different directions of object movements. Moreover, we show that the position information predicts the input's future position taking into account its moving direction due to the direction-selective responses of the dorsal-like units.

UR - http://www.scopus.com/inward/record.url?scp=84867685631&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84867685631&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-33269-2_68

DO - 10.1007/978-3-642-33269-2_68

M3 - Conference contribution

AN - SCOPUS:84867685631

SN - 9783642332685

VL - 7552 LNCS

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

SP - 539

EP - 546

BT - Artificial Neural Networks and Machine Learning, ICANN 2012 - 22nd International Conference on Artificial Neural Networks, Proceedings

ER -