Title :
A learning algorithm for MLN with dynamic neurons
Author :
Li, Haizhou ; Xu, Bingzheng
Author_Institution :
Inst. of Radio Autom., South China Univ. of Technol., Guangzhou, China
Abstract :
A multilayer network architecture with dynamic neurons which have multilocal feedbacks is built. The proposed network can be trained to memorize sequential patterns. A learning algorithm which is more effective and easier to implement is derived. Some experiments on speech recognition of Chinese digits designed to explore the capabilities of the proposed networks to learn dynamic properties of time-varying data are discussed. The performance of dynamic neurons with different time delay periods is also shown
Keywords :
learning systems; neural nets; parallel architectures; speech recognition; Chinese digits; dynamic neurons; feedbacks; learning algorithm; multilayer network architecture; sequential pattern memorization; speech recognition; time-varying data dynamic properties learning; Added delay; Automation; Delay effects; Hysteresis; Neural networks; Neurofeedback; Neurons; Nonhomogeneous media; Speech recognition; State feedback;
Conference_Titel :
Neural Networks, 1991. 1991 IEEE International Joint Conference on
Print_ISBN :
0-7803-0227-3
DOI :
10.1109/IJCNN.1991.170453
