open access

https://doi.org/10.14943/doctoral.k16190

https://hdl.handle.net/2115/94062

Z6-Ⅱオービフォールドコンパクト化におけるE8×E8ヘテロティック弦理論でのMSSM様モデルの生成AIによる探索に関する研究

英語

2024-12-25

Hokkaido University

63p

String theory is a leading candidate for unifying gravity with quantum mechanics, avoiding quantum anomalies. Due to its ten-dimensional nature, six extra dimensions must be compactified to unobservable scales to be consistent with our four-dimensional reality. This compactification process leads to a vast landscape of possible four-dimensional string models, estimated to be on the order of 10 1500. Among these, MSSM-like models, which resemble the Minimal Supersymmetric Standard Model, are of particular interest but represent only a small fraction of the landscape. Traditional random scanning methods are often employed to search for these MSSM-like models within the string landscape. However, the relationship between compactification parameters and resulting spectra is complex, making this process computationally challenging. Recent research has proposed using autoencoders (AE) and Variational Autoencoders (VAE) to map the high-dimensional parameter space onto a lower-dimensional space, improving the search efficiency for MSSM-like models. This dissertation explores the use of VAE-based generative models to generate new MSSMlike data, demonstrating their potential to efficiently explore the string landscape. We also introduce the “phased data augmentation” technique to enhance the training of these models on sparse datasets. The effectiveness of this approach is evaluated through experiments, with results showing promising advancements in the search for viable MSSM-like models.

2024-12-25

甲第16190号

博士(理学)

(主査) 教授 小林 達夫, 教授 鈴木 久男, 准教授 瀬戸 治, 講師 末廣 一彦

理学院（宇宙理学専攻）

博士論文

Version of Record