TY - GEN
T1 - Encoding Multisensory Information in Modular Neural Networks
AU - Wang, He
AU - Zhang, Wen Hao
AU - Wong, K. Y.Michael
AU - Wu, Si
N1 - Funding Information:
Acknowledgments. 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:
© 2017, Springer International Publishing AG.
PY - 2017
Y1 - 2017
N2 - The brain is capable of integrating information in multiple sensory channels in a Bayesian optimal way. Based on a decentralized network model inspired by electrophysiological recordings, we consider the structural pre-requisites for optimal multisensory integration. In this architecture, same-channel feedforward and recurrent links encode the unisensory likelihoods, whereas reciprocal couplings connecting the different modules are shaped by the correlation in the joint prior probabilities. Moreover, the statistical relationship between the difference in the optimal network structures and the difference in the priors and the likelihoods clearly shows that the network can encode multisensory information in a distributed manner. Our results generate testable predictions for future experiments and are likely to be applicable to other artificial systems.
AB - The brain is capable of integrating information in multiple sensory channels in a Bayesian optimal way. Based on a decentralized network model inspired by electrophysiological recordings, we consider the structural pre-requisites for optimal multisensory integration. In this architecture, same-channel feedforward and recurrent links encode the unisensory likelihoods, whereas reciprocal couplings connecting the different modules are shaped by the correlation in the joint prior probabilities. Moreover, the statistical relationship between the difference in the optimal network structures and the difference in the priors and the likelihoods clearly shows that the network can encode multisensory information in a distributed manner. Our results generate testable predictions for future experiments and are likely to be applicable to other artificial systems.
KW - Multisensory Bayesian inference
KW - Recurrent neural networks
UR - http://www.scopus.com/inward/record.url?scp=85035118421&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85035118421&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-70093-9_70
DO - 10.1007/978-3-319-70093-9_70
M3 - Conference contribution
AN - SCOPUS:85035118421
SN - 9783319700922
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 658
EP - 665
BT - Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings
A2 - Liu, Derong
A2 - Xie, Shengli
A2 - Li, Yuanqing
A2 - El-Alfy, El-Sayed M.
A2 - Zhao, Dongbin
PB - Springer Verlag
T2 - 24th International Conference on Neural Information Processing, ICONIP 2017
Y2 - 14 November 2017 through 18 November 2017
ER -