next up previous contents
Next: 目次

概要:

In recent years, neural networks began to be utilized in many fields. Originally, the system was used for logical operation, and later for recognition of pattern and character, voice and so on. It is also applied for the estimate of weather and stock price or control of machinery.

The neural network is one which imitated the nervous system in the brain, and a chaotic neuron is modeled after the nerve cell.

In this study, applying the back-propagation learning method to chaotic parameters in a chaotic neuron, and whether this method is effective or not is examined. A few of periodic patterns is memorized to a 3-layer perseptron using chaotic neurons. Then, when optional pattern is input into the network, the detection of memorized patterns is performed.

It is unable in principle for chaotic neural networks to memorize past input pattern in order. Because of that, this method isn't suitable for the exact detection, but it is expected that it can apply for the statistical detection. Some effectiveness is shown through the experiment.





出口研究室へ

Deguchi Toshinori
1998年04月01日 (水) 12時03分23秒 JST