next up previous contents
Next: 目次   目次

概要:

The learning ability and the memory ability of the human brain have the excellent processing performance that the present computers can't achieve. The neural network is a technology that reproduces the human brain artificially.

The hierarchical neural network that has the internal memory layer is suitable for learning of time-series.

In my laboratory, the ``delayed learning method'' was newly devised. Using this method, the neural network with internal memory can learn complicated time-series.

In this study, the influence on the learning result of ``delay time'' and ``number of elements of network'' are examined. ``Delay time'' is one of the parameters of the delayed learning method. The network composed of several elements tried learning many kinds of time-series. And the difference of a learning result was compared and discussed.

A sine wave, a triangular wave, a saw tooth wave, a reverse-saw tooth wave and the pulse wave were used for the time-series. The experimental results showed that the number of elements of network has a big influence on deciding effective delay time.

On the number of middle elements, it was confirmed that the number of learning successes reached the ceiling by a certain number of elements.

On the number of internal elements, it was confirmed it doesn't have to be a large number to succeed of learning.

It was confirmed that these results were somewhat different according to the time-series.




next up previous contents
Next: 目次   目次
Deguchi Lab. 2011年3月3日