Title :
A complex sequence recognition model
Author :
Tanaka, Takehisa
Author_Institution :
Center for Neural Eng., Univ. of Southern California, Los Angeles, CA, USA
Abstract :
The author proposes a biologically plausible model to learn and recognize complex sequences. He uses decreasing outputs and units which order recurring symbols. Though the model does not deal with complicated details of biological neurons, it is more biologically plausible and easier to implement on real hardware than other models. It also avoids crosstalk of memorized sequences. An arbitrary number and length of sequences can be recognized. Learning of sequences is simple and it is possible to preset weights analytically. Time periods of presenting sequences and symbols do not affect recognition
Keywords :
neural nets; pattern recognition; time series; complex sequence recognition model; memorized sequences; presentation time periods; recurring symbols; temporal behaviour; temporal patterns; Abstracts; Biological system modeling; Crosstalk; Delay effects; Detectors; Equations; Hardware; Mathematical model; Neurofeedback; Neurons;
Conference_Titel :
Neural Networks, 1992. IJCNN., International Joint Conference on
Conference_Location :
Baltimore, MD
Print_ISBN :
0-7803-0559-0
DOI :
10.1109/IJCNN.1992.227341
