How the prior information shapes neural networks for optimal multisensory integration

He Wang; Wen Hao Zhang; K. Y.Michael Wong; Si Wu

doi:10.1007/978-3-319-59081-3_16

How the prior information shapes neural networks for optimal multisensory integration

He Wang, Wen Hao Zhang, K. Y.Michael Wong, Si Wu

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

3 Scopus citations

Abstract

Extensive studies suggest that the brain integrates multisensory signals in a Bayesian optimal way. In this work, we consider how the couplings in a neural network model are shaped by the prior information when it performs optimal multisensory integration and encodes the whole profile of the posterior. To process stimuli of two modalities, a biologically plausible neural network model consists of two modules, one for each modality, and crosstalks between the two modules are carried out through feedforward cross-links and reciprocal connections. We found that the reciprocal couplings are crucial to optimal multisensory integration in that their pattern is shaped by the correlation in the joint prior distribution of sensory stimuli. Our results show that a decentralized architecture based on reciprocal connections is able to accommodate complex correlation structures across modalities and utilize this prior information in optimal multisensory integration.

Original language	English (US)
Title of host publication	Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings
Editors	Fengyu Cong, Qinglai Wei, Andrew Leung
Publisher	Springer Verlag
Pages	128-136
Number of pages	9
ISBN (Print)	9783319590806
DOIs	https://doi.org/10.1007/978-3-319-59081-3_16
State	Published - 2017
Externally published	Yes
Event	14th International Symposium on Neural Networks, ISNN 2017 - Sapporo, Hakodate, and Muroran, Hokkaido, Japan Duration: Jun 21 2017 → Jun 26 2017

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10262 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	14th International Symposium on Neural Networks, ISNN 2017
Country/Territory	Japan
City	Sapporo, Hakodate, and Muroran, Hokkaido
Period	6/21/17 → 6/26/17

Keywords

Bayesian inference
Multisensory processing
Recurrent neural networks

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-319-59081-3_16

Cite this

Wang, H., Zhang, W. H., Wong, K. Y. M., & Wu, S. (2017). How the prior information shapes neural networks for optimal multisensory integration. In F. Cong, Q. Wei, & A. Leung (Eds.), Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings (pp. 128-136). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10262 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-59081-3_16

How the prior information shapes neural networks for optimal multisensory integration. / Wang, He; Zhang, Wen Hao; Wong, K. Y.Michael et al.
Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings. ed. / Fengyu Cong; Qinglai Wei; Andrew Leung. Springer Verlag, 2017. p. 128-136 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10262 LNCS).

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

Wang, H, Zhang, WH, Wong, KYM & Wu, S 2017, How the prior information shapes neural networks for optimal multisensory integration. in F Cong, Q Wei & A Leung (eds), Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10262 LNCS, Springer Verlag, pp. 128-136, 14th International Symposium on Neural Networks, ISNN 2017, Sapporo, Hakodate, and Muroran, Hokkaido, Japan, 6/21/17. https://doi.org/10.1007/978-3-319-59081-3_16

Wang H, Zhang WH, Wong KYM, Wu S. How the prior information shapes neural networks for optimal multisensory integration. In Cong F, Wei Q, Leung A, editors, Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings. Springer Verlag. 2017. p. 128-136. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-59081-3_16

Wang, He ; Zhang, Wen Hao ; Wong, K. Y.Michael et al. / How the prior information shapes neural networks for optimal multisensory integration. Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings. editor / Fengyu Cong ; Qinglai Wei ; Andrew Leung. Springer Verlag, 2017. pp. 128-136 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{def18fe925c843b38b9286ffe823aa8a,

title = "How the prior information shapes neural networks for optimal multisensory integration",

abstract = "Extensive studies suggest that the brain integrates multisensory signals in a Bayesian optimal way. In this work, we consider how the couplings in a neural network model are shaped by the prior information when it performs optimal multisensory integration and encodes the whole profile of the posterior. To process stimuli of two modalities, a biologically plausible neural network model consists of two modules, one for each modality, and crosstalks between the two modules are carried out through feedforward cross-links and reciprocal connections. We found that the reciprocal couplings are crucial to optimal multisensory integration in that their pattern is shaped by the correlation in the joint prior distribution of sensory stimuli. Our results show that a decentralized architecture based on reciprocal connections is able to accommodate complex correlation structures across modalities and utilize this prior information in optimal multisensory integration.",

keywords = "Bayesian inference, Multisensory processing, Recurrent neural networks",

author = "He Wang and Zhang, {Wen Hao} and Wong, {K. Y.Michael} and Si Wu",

note = "Funding Information: This work is supported by the Research Grants Council of Hong Kong (N_HKUST606/12, 605813 and 16322616) and National Basic Research Program of China (2014CB846101) and the Natural Science Foundation of China (31261160495). Publisher Copyright: {\textcopyright} Springer International Publishing AG 2017.; 14th International Symposium on Neural Networks, ISNN 2017 ; Conference date: 21-06-2017 Through 26-06-2017",

year = "2017",

doi = "10.1007/978-3-319-59081-3_16",

language = "English (US)",

isbn = "9783319590806",

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

publisher = "Springer Verlag",

pages = "128--136",

editor = "Fengyu Cong and Qinglai Wei and Andrew Leung",

booktitle = "Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings",

}

TY - GEN

T1 - How the prior information shapes neural networks for optimal multisensory integration

AU - Wang, He

AU - Zhang, Wen Hao

AU - Wong, K. Y.Michael

AU - Wu, Si

N1 - Funding Information: This work is supported by the Research Grants Council of Hong Kong (N_HKUST606/12, 605813 and 16322616) and National Basic Research Program of China (2014CB846101) and the Natural Science Foundation of China (31261160495). Publisher Copyright: © Springer International Publishing AG 2017.

PY - 2017

Y1 - 2017

N2 - Extensive studies suggest that the brain integrates multisensory signals in a Bayesian optimal way. In this work, we consider how the couplings in a neural network model are shaped by the prior information when it performs optimal multisensory integration and encodes the whole profile of the posterior. To process stimuli of two modalities, a biologically plausible neural network model consists of two modules, one for each modality, and crosstalks between the two modules are carried out through feedforward cross-links and reciprocal connections. We found that the reciprocal couplings are crucial to optimal multisensory integration in that their pattern is shaped by the correlation in the joint prior distribution of sensory stimuli. Our results show that a decentralized architecture based on reciprocal connections is able to accommodate complex correlation structures across modalities and utilize this prior information in optimal multisensory integration.

AB - Extensive studies suggest that the brain integrates multisensory signals in a Bayesian optimal way. In this work, we consider how the couplings in a neural network model are shaped by the prior information when it performs optimal multisensory integration and encodes the whole profile of the posterior. To process stimuli of two modalities, a biologically plausible neural network model consists of two modules, one for each modality, and crosstalks between the two modules are carried out through feedforward cross-links and reciprocal connections. We found that the reciprocal couplings are crucial to optimal multisensory integration in that their pattern is shaped by the correlation in the joint prior distribution of sensory stimuli. Our results show that a decentralized architecture based on reciprocal connections is able to accommodate complex correlation structures across modalities and utilize this prior information in optimal multisensory integration.

KW - Bayesian inference

KW - Multisensory processing

KW - Recurrent neural networks

UR - http://www.scopus.com/inward/record.url?scp=85021710146&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85021710146&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-59081-3_16

DO - 10.1007/978-3-319-59081-3_16

M3 - Conference contribution

AN - SCOPUS:85021710146

SN - 9783319590806

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

SP - 128

EP - 136

BT - Advances in Neural Networks - ISNN 2017 - 14th International Symposium, ISNN 2017, Proceedings

A2 - Cong, Fengyu

A2 - Wei, Qinglai

A2 - Leung, Andrew

PB - Springer Verlag

T2 - 14th International Symposium on Neural Networks, ISNN 2017

Y2 - 21 June 2017 through 26 June 2017

ER -

How the prior information shapes neural networks for optimal multisensory integration

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this