Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/71439
Title: | 高粒度量能器測試粒子且利用深度學習來分辨粒子 High Granularity Calorimeter(HGCAL)Beam Test Particle Separation with Deep Learning |
Authors: | Wen-Liang Huang 黃文亮 |
Advisor: | 陳凱風(Kai-Feng Chen) |
Keyword: | 高能物理,量能器,量能方式,能量重建,機器學習,深度學習,卷機神經網路, High energy physics,Calorimeter,Calorimeter method,Energy reconstructor,Machine learning,Deep learning,Convolution neural network, |
Publication Year : | 2020 |
Degree: | 碩士 |
Abstract: | 為了大型強子對撞機中原本已經很高的光度更增加十倍以上來讓對撞 且發散的粒子能夠更多被留下來,高粒度量熱器 (HGCAL) 就是在緊湊緲子 螺旋 (CMS) 中裡第二階段中想要製作出新的測量器。此次實驗中只討論高 粒度量熱器中的電子量熱器 (EE)。為了驗證在光束測量中產生的資料和模 擬粒子可以視為他們的能量為變數,28 層的能量可以視為 28 個變數,而這 28 個變數來利用傳統的機器學習和最近熱門中深度學習裡的卷積神經網路 (Convolution Neural Network) 來確認兩個粒子他們分離的機率,以及資料跟 模型他們的相似程度。比較在機器學習和卷積神經網路演算法訓練結果,後 者為了分離電子跟 π 介子其準確率比前者還更大。在之後可以在卷積神經網 路程式館裡建立專有的六角形模組,並比較和之前的準確率來確認六角形模 組訓練是不是真的比較好 Large Hadron Collider(LHC) is processing to the High Luminosity phase, it will deliver 10 times more integrated luminosity than now. HighGranularity Calorime ter(HGCAL) is the chosen technology by the Compact Muon Solenoid(CMS) ex periment as part of the phase 2 upgrading program.This experiment will focus on the electromagnetic calorimter(EE)region. The already generated data and corresponded simulation can look upon their energies as the variables in each layer of electromag netic calorimter.Then using machine learning algorithms and Convolution Neural Network(CNN) algorithm to check that the input and output can correspond to each other and their probability of the data and simulation.Comapring the machine learn ing algorithm result to the CNN algorithm result that the latter one can have a higher accuracy to separate e+ and π+ than the before one. Outlook is to construct a hexag onal image module in the CNN library and compare the accuracy is higher than the CNN in this thesis. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/71439 |
DOI: | 10.6342/NTU202004357 |
Fulltext Rights: | 有償授權 |
Appears in Collections: | 應用物理研究所 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
U0001-2511202017534500.pdf Restricted Access | 9.32 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.