神經網路光學電腦;All Optical Neural Networks

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/75175

Title:	神經網路光學電腦;All Optical Neural Networks
Authors:	周達業;Chou, Da-Ya
Contributors:	光電科學與工程學系
Keywords:	光學電腦;訊號辨識;類神經網路;All optical computing;Signal recognition;Artificial neural networks
Date:	2017-07-07
Issue Date:	2017-10-27 17:17:27 (UTC+8)
Publisher:	國立中央大學
Abstract:	本論文主要目的是透過光通訊模擬軟體來設計出Echo State Network(ESN)之結構，在結構中的類神經元都是由光纖、兩個耦合器和摻鉺光纖放大器所組成，並用於處理與時間有關的數據，使其能夠分辨訊號。本論文將改變類神經元接法、數目，或者輸入的光源，並以NRMSE來當作辨別訊號能力的好壞，來找尋一個最好的結構。;The main purpose of this paper is to design the structure of Echo State Network (ESN) through optical communication simulation software. Our structure uses several neurons. The neuron is composed of optical fibers, two couplers and erbium-doped fiber amplifiers (EDFA). We use structure to deal with time-related data, so that it can recognize the signal. This paper will change the connection of neurons, the number of neurons, or the input light source. We use NRMSE to identify the ability of the signal is good or bad, to find a best structure.
Appears in Collections:	[Graduate Institute of Optics and Photonics] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	194	View/Open

社群 sharing