Presentation	2004/3/12 Discretization of analog communication signals by noise addition in reinforcement learning of communication Katsunari SHIBATA,
PDF Download Page	PDF download Page Link
Abstract(in Japanese)	(See Japanese page)
Abstract(in English)	Towards the unified processing of symbols and patterns by neural networks, it was examined that symbols emerge using neural networks that is trained only by reinforcement learning. A very simple communication-learning task was assumed, and some noise is added to the communication signals. After learning, as the noise level during learning became larger, the communication signals were binarized more, and the system became more tolerant of noise unless the noise level was too large. The receiver was also trying to interpret the signals as binarized value. Furthermore, it was examined that recurrent neural networks promote the discretization.
Keyword(in Japanese)	(See Japanese page)
Keyword(in English)	reinforcement learning / neural network / communication learning / discretization of signals / symbol emergence
Paper #	NC2003-203
Date of Issue

Conference Information
Committee	NC
Conference Date	2004/3/12(1days)
Place (in Japanese)	(See Japanese page)
Place (in English)
Topics (in Japanese)	(See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To	Neurocomputing (NC)
Language	JPN
Title (in Japanese)	(See Japanese page)
Sub Title (in Japanese)	(See Japanese page)
Title (in English)	Discretization of analog communication signals by noise addition in reinforcement learning of communication
Sub Title (in English)
Keyword(1)	reinforcement learning
Keyword(2)	neural network
Keyword(3)	communication learning
Keyword(4)	discretization of signals
Keyword(5)	symbol emergence
1st Author's Name	Katsunari SHIBATA
1st Author's Affiliation	Dept. of Electrical & Electronic Engineering, Oita University()
Date	2004/3/12
Paper #	NC2003-203
Volume (vol)	vol.103
Number (no)	734
Page	pp.pp.-
#Pages	6
Date of Issue