Multiplex Communication in 2D Mesh Neural Network
Lists
Only you see this
Abstract
We have already analyzed and reported on multiplex communication in neural network. However, these are composed and designed so as to force the network communicate with multiplexed manner using "code" or "temporal sequence." That is, the network has a main loop and some coding/decoding circuits, and these are somewhat artificial. In this paper, we show without these artificial guidance, it is also possible to communicate with multiplexing in a natural 2D grid shape neural network, where some learning algorithm are employed to facilitate to find paths from a transmitting neuron to a receiving neuron. We also show that in these neural networks, random sequences are more frequent in number than that of non-random sequences.
BibTeX
@inproceedings{kamimura2015multiplex,
author = {Takuya Kamimura and Yasushi Yagi and Shinichi Tamura and Y.W Chen},
title = {{Multiplex Communication in 2D Mesh Neural Network}},
booktitle = {Advances},
year = {2015},
doi = {10.2991/cisia-15.2015.157},
}