請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/8398
標題: | 利用深度學習辨認強子對撞機產生之W’玻色子信號 Distinguishing W’ Signals at Hadron Colliders Using Deep Learning |
作者: | Ting-Kuo Chen 陳定國 |
指導教授: | 蔣正偉(Cheng-Wei Chiang) |
關鍵字: | 強子對撞機,W'玻色子,深度學習, Deep Learning,W' Boson,Hadron Colliders, |
出版年 : | 2020 |
學位: | 碩士 |
摘要: | 我們提出一個以深度學習為基礎、透過分析其衰變到帶電輕子及橫向遺失動能末態特性的方法,來分辨各種不同向量及純量單電荷玻色子的作用機制──而這些作用機制傳統上在強子對撞機中會因為一個四重的模糊性變得難以分析。我們提出幾種不同的深度學習模式來處理這個問題:其一是直接使用可見的帶電輕子運動學可觀測量來訓練一個深度類神經網路;其二是利用帶電輕子的橫動量及膺快度來製成二維直方圖,並以其訓練一個卷積類神經網路及一個移轉學習網路。經研究後我們發現卷積類神經網路的表現遠比深度類神經網路及移轉學習網路優異。我們也分析了這個方法在不同信號顯著度的狀況下,以無關模型之基礎應用的可靠度。最後,我們也演示了該方法如何被更廣泛地應用在帶有強子噴流末態的次領導對撞過程中。這篇文章所探討的方法將可以為未來新發現的超越標準模型玻色子特性之分析帶來幫助。 We demonstrate a neural-network (NN)-based scheme to distinguish the couplings of different hypothesis W' through the lepton + missing transverse energy channel, which is traditionally challenging due to a four-fold ambiguity at hadron colliders, such as the Large Hadron Collider (LHC). We also take hypothesis scalar charged bosons into consideration. In addition to constructing a simple dense neural network (DNN) trained upon the kinematic information of the visible charged lepton, we further construct a convolutional neural network (CNN) and a transfer-learning network (TLN) trained upon 2D histograms made from the transverse momentum and pseudorapidity of the charged lepton, and conclude that the CNN outperforms both the DNN and the TLN. We also investigate the reliability of a model-independent application of this method to different significance scenarios. In addition, we compare our method with a few traditional hypothesis tests and discuss the pros and cons. Finally, we demonstrate a more general application to the next-to-leading-order (NLO) process in which an additional final-state jet is present. The scheme presented in this paper could serve to help investigate the properties of newly discovered W' in the future. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/8398 |
DOI: | 10.6342/NTU202001854 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 物理學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
U0001-2507202023010500.pdf | 4.34 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。