TY - GEN

T1 - Learning and recognition of multiple fluctuating temporal patterns using S-CTRNN

AU - Murata, Shingo

AU - Arie, Hiroaki

AU - Ogata, Tetsuya

AU - Tani, Jun

AU - Sugano, Shigeki

PY - 2014

Y1 - 2014

N2 - In the present study, we demonstrate the learning and recognition capabilities of our recently proposed recurrent neural network (RNN) model called stochastic continuous-time RNN (S-CTRNN). S-CTRNN can learn to predict not only the mean but also the variance of the next state of the learning targets. The network parameters consisting of weights, biases, and initial states of context neurons are optimized through maximum likelihood estimation (MLE) using the gradient descent method. First, we clarify the essential difference between the learning capabilities of conventional CTRNN and S-CTRNN by analyzing the results of a numerical experiment in which multiple fluctuating temporal patterns were used as training data, where the variance of the Gaussian noise varied among the patterns. Furthermore, we also show that the trained S-CTRNN can recognize given fluctuating patterns by inferring the initial states that can reproduce the patterns through the same MLE scheme as that used for network training.

AB - In the present study, we demonstrate the learning and recognition capabilities of our recently proposed recurrent neural network (RNN) model called stochastic continuous-time RNN (S-CTRNN). S-CTRNN can learn to predict not only the mean but also the variance of the next state of the learning targets. The network parameters consisting of weights, biases, and initial states of context neurons are optimized through maximum likelihood estimation (MLE) using the gradient descent method. First, we clarify the essential difference between the learning capabilities of conventional CTRNN and S-CTRNN by analyzing the results of a numerical experiment in which multiple fluctuating temporal patterns were used as training data, where the variance of the Gaussian noise varied among the patterns. Furthermore, we also show that the trained S-CTRNN can recognize given fluctuating patterns by inferring the initial states that can reproduce the patterns through the same MLE scheme as that used for network training.

KW - S-CTRNN

KW - recurrent neural network

KW - variance estimation

UR - http://www.scopus.com/inward/record.url?scp=84958521901&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84958521901&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-11179-7_2

DO - 10.1007/978-3-319-11179-7_2

M3 - Conference contribution

AN - SCOPUS:84958521901

SN - 9783319111780

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

SP - 9

EP - 16

BT - Artificial Neural Networks and Machine Learning, ICANN 2014 - 24th International Conference on Artificial Neural Networks, Proceedings

PB - Springer Verlag

T2 - 24th International Conference on Artificial Neural Networks, ICANN 2014

Y2 - 15 September 2014 through 19 September 2014

ER -