Deep autoencoders for ATLAS data compression

Storage is one of the main limiting factors to the recording of information from proton-proton collision events at the Large Hadron Collider (LHC), at CERN in Geneva. Hence, the ATLAS experiment at the LHC uses a trigger system, which selects and transfers interesting events to the data storage system. However, if those events are buried in large backgrounds and difficult to identify as a signal by the trigger system, they will also be discarded together with the background. To alleviate this problem, various compression algorithms are proposed to reduce the size of the data that is recorded. One of those algorithms is an autoencoder (AE) network that tries to implement an approximation to the identity, f(x) = x. Its goal is to create a lower-dimensional representation of the input data in a latent space. Then using this latent representation the model can reconstruct its input. Moreover, two advanced families of AEs are proposed, the Variational and the Adversarial AE. Next, various combinations of different compression algorithms are tested to compress and construct accurate representations of the input data. Finally, the studied AEs are implemented for anomaly detection.