Jawahar Pratik, Aarrestad Thea, Chernyavskaya Nadezda, Pierini Maurizio, Wozniak Kinga A, Ngadiuba Jennifer, Duarte Javier, Tsan Steven
Experimental Physics Department, European Center for Nuclear Research (CERN), Geneva, Switzerland.
Faculty of Computer Science, University of Vienna, Vienna, Austria.
Front Big Data. 2022 Feb 28;5:803685. doi: 10.3389/fdata.2022.803685. eCollection 2022.
We investigate how to improve new physics detection strategies exploiting variational autoencoders and normalizing flows for anomaly detection at the Large Hadron Collider. As a working example, we consider the DarkMachines challenge dataset. We show how different design choices (e.g., event representations, anomaly score definitions, network architectures) affect the result on specific benchmark new physics models. Once a baseline is established, we discuss how to improve the anomaly detection accuracy by exploiting normalizing flow layers in the latent space of the variational autoencoder.
我们研究如何利用变分自编码器和归一化流来改进新物理探测策略，以便在大型强子对撞机上进行异常检测。作为一个实际例子，我们考虑DarkMachines挑战数据集。我们展示了不同的设计选择（例如，事件表示、异常分数定义、网络架构）如何影响特定基准新物理模型的结果。一旦建立了基线，我们将讨论如何通过在变分自编码器的潜在空间中利用归一化流层来提高异常检测的准确性。
