Exploring Anomalous Quartic Gauge Couplings with Machine Learning
Delve into the study of anomalous quartic gauge couplings (aQGCs) using machine learning algorithms at Liaoning Normal University. Discover the significance of aQGCs in high-energy physics research, including their impact on VBS processes and extraction of energy in collider experiments. Explore the application of isolation forest and neural networks to analyze signals related to aQGCs and nTGCs, shedding light on the complexities of high-energy physics phenomena. Uncover the role of artificial neural networks in this study, aiming to avoid overfitting and ensuring accurate results through extensive training and validation datasets containing around 600,000 events.
Download Presentation
Please find below an Image/Link to download the presentation.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author. If you encounter any issues during the download, it is possible that the publisher has removed the file from their server.
You are allowed to download the files provided on this website for personal or commercial use, subject to the condition that they are used lawfully. All files are the property of their respective owners.
The content on the website is provided AS IS for your information and personal use only. It may not be sold, licensed, or shared on other websites without obtaining consent from the author.
E N D
Presentation Transcript
Use machine learning to Use machine learning to study study aQGCs aQGCs Ji-Chong Yang Liaoning Normal University Collaborator Yu-Chen Guo, Chong-Xing Yue 2021-11-9 CEPC 2021
Machine learning algorithms has been widely used in HEP: 1/30
Outline Extract energy of VV to VV subprocess in VBS using neural network Using isolation forest to study the signal of aQGCs Using modified isolation forest to study the signal of nTGCs at e+ e- collider (we are still working on this). 2/30
Anomalous quartic gauge couplings Dimension-8 SMEFT operators Many reasons to study aQGCs aTGCs induced by loop aQGCs induced by tree level Important in operator space Appear in theories that assume a finite electromagnetic field strength 3/30
VBS and aQGCs Introduce the arena 18 variables 4/30
Energy is an important parameter in the study of an EFT 6/30
When digging for information one must ensure that there is indeed information to dig (1) x+y=0 (2) x+y=0, and: 7/30
approximation 8/30
Avoid over fit 10/30
Results Training dataset and Validation dataset: About 600,000 events 11/30
What information is used 12/30
What we can do with ANN find more understandable patterns 9075 fitting variables -> 5 13/30
What we can do with ANN Highlight signals Unitarily bound 14/30
Simplify the phenomenological study of SMEFT need to spend a lot of time on kinematic analysis 15/30
Event selection strategy (1) SM is successful the signal of NP is few. (2) In SMEFT, the signal events are induced by new interactions. Signal events are kinematically different. Event selection strategy. 16/30
A common feature of many NP model Few Different 17/30
Anomaly detection Many machine learning algorithms on anomaly detection Auto-encode Isolation forest 18/30
Isolation tree Randomly choose a undivided leaf Randomly choose a dimension Divide 19/30
anomaly points are easier to isolate 0 0 0 01 01 01 01 01 0 0 0 00 00 00 00 00 1 10 10 10 100 100 100 100 1 10 10 10 101 1010 1010 10100 1 11 110 110 110 110 1101 1101 1 11 110 110 110 110 1100 1100 1 10 10 10 101 1011 1011 1011 1 11 111 111 111 111 111 111 1 10 10 10 101 1010 1010 10101 20/30
Tree -> forest Isolation tree: Isolation forest: 21/30
Result: 22/30
Advantage of IF Transparent: just quantify anomaly Independent of the operators (automatic event selection strategy) The anomaly events are important even when NP is absent 23/30
Disadvantage How about when the signal events are few but NOT very different? For example, interference, which is important when Wilson coefficients are sufficiently small. 24/30
The problem 25/30
sacrificed some of generality (but not all). ---- MC data has to be generated according to the operator to be studied. ---- However, the procedure is independent of operators to be studied. (still automatic event selection strategy) We focus on the change of anomaly score (still unsupervised training, no need to tag the training data which is impossible for interferences) 27/30
The strategy Training data set: MC data with SM A reference anomaly score distribution is built on SM data. Data set to be investigated: MC data with different coefficients (including zero coefficient) relate the points in this data set to the point of training data set (by distance), calculate the difference of anomaly scores. 28/30
Result Traditional cuts:2102.03623 MIF: 29/30
Summary ANN can be used to study the energy of subprocess AA->WW in VBS. (2107.13624). Isolation forest can be used to highlight the signal events of aQGCs. (2103.03151) A modified isolation forest algorithm can be useful to study the nTGCs at the CEPC. (will be on arXiv soon). Many has been done, more to do Thank you! 30/30
