TY - GEN
T1 - Encoding Longer-term Contextual Sensorimotor Information in a Predictive Coding Model
AU - Zhong, Junpei
AU - Ogata, Tetsuya
AU - Cangelosi, Angelo
N1 - Funding Information:
ACKNOWLEDGEMENT The research was partially supported by New Energy and Industrial Technology Development Organization (NEDO). A Pytorch implementation of MT-AFA-PredNet can be found on Github2
Publisher Copyright:
© 2018 IEEE.
PY - 2019/1/28
Y1 - 2019/1/28
N2 - Studies suggest that the difference of the sensorimotor events can be recorded with the fast- and slower-changing neural activities in the hierarchical brain areas, in which they have bi-directional connections. The slow-changing representations attempt to predict the activities on the faster level by relaying categorized sensorimotor events. On the other hand, the incoming sensory information corrects such event-based prediction on the higher level by the novel or surprising signal. From this motivation, we propose a predictive hierarchical artificial neural network model which is implemented the differentiated temporal parameters for neural updates. Also, both the fast- and slow-changing neural activities are modulated by the active motor activities. This model is examined in the driving dataset, recorded in various events, which incorporates the image sequences and the ego-motion of the vehicle. Experiments show that the model encodes the driving scenarios on the higher-level where the neuron recorded the long-term context.
AB - Studies suggest that the difference of the sensorimotor events can be recorded with the fast- and slower-changing neural activities in the hierarchical brain areas, in which they have bi-directional connections. The slow-changing representations attempt to predict the activities on the faster level by relaying categorized sensorimotor events. On the other hand, the incoming sensory information corrects such event-based prediction on the higher level by the novel or surprising signal. From this motivation, we propose a predictive hierarchical artificial neural network model which is implemented the differentiated temporal parameters for neural updates. Also, both the fast- and slow-changing neural activities are modulated by the active motor activities. This model is examined in the driving dataset, recorded in various events, which incorporates the image sequences and the ego-motion of the vehicle. Experiments show that the model encodes the driving scenarios on the higher-level where the neuron recorded the long-term context.
UR - http://www.scopus.com/inward/record.url?scp=85062767836&partnerID=8YFLogxK
U2 - 10.1109/SSCI.2018.8628911
DO - 10.1109/SSCI.2018.8628911
M3 - Conference article published in proceeding or book
AN - SCOPUS:85062767836
T3 - Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018
SP - 160
EP - 167
BT - Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence, SSCI 2018
A2 - Sundaram, Suresh
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 8th IEEE Symposium Series on Computational Intelligence, SSCI 2018
Y2 - 18 November 2018 through 21 November 2018
ER -