Generating events with GANs

In the coming years, the LHC experiments will need to produce an incredible amount of Monte Carlo events to match the expected increase in luminosity. This challenge, together with the need of using more sophisticated generators, will stretch the computing resources and may limit the physics reach of the experiments.

In our paper, we provide a solution to overcome these problems by using Machine Learning techniques, specifically Generative Adversarial Networks (GANs). In this first attempt, we focussed on the relatively simple di-jet events but we also prepared the tools to produce more complex events such as top quark pair events and multi boson events which are produced in large quantities at LHC.

We actually trained two networks, one using the output of the generators (which is commonly called particle level) and another after the detector simulation and reconstructions (also known as reconstruction or detector level). The good agreement of first network with the training sample demonstrates that it is possible to use a relatively small number of events to train a GAN which can be subsequently be used to significantly increase the number of events as we are able to to generate 1 million events in less than a minute.  The good agreement of second network demonstrate that the detector response and even the reconstruction steps can be reproduced by a GAN.

The results we obtained are a clear indication that GANs can have a high impact on LHC experiment in several areas. You can find more details on the paper but we would like to share a video showing the learning of the GAN as it is trained which cannot be included in the paper.

Let us know what you think of this new application of the GAN in the comments below or by contacting the authors of the paper.

Michele and Serena

 

DiJetGAN_animation.gif

University of Edinburgh

Welcome to the home of the Higgs boson inventor and Nobel prize winner Peter Higgs.

Edinburgh is the capital of Scotland and a very nice place to live, especially if you prefer rugby over football and you will find the Scots to be friendly and welcoming, although sometimes you may struggle to understand them. The city has all services you may need including theatres, cinemas, independent restaurants, and much more. The public transport will take you anywhere including to the airport that connects the city to many major cities in Europe and beyond. We even have a direct flight to Geneva!

The University is a major player in the city recent development making a strong effort in creating new connections between academia and industries. The recently open Bayes Centre is a new building in the town campus that will enhance the capabilities of the University to develop multidisciplinary projects centred on the use of advanced statistical tools in a variety of fields. Why does this sound familiar? Ah, yes, that’s what we do!

The University has several campuses as it could no longer be hosted in the city centre.  Actually, the School of Physics and Astronomy is based in the King’s Building campus which is located 20 minutes by bus from the city centre. The campus hosts many schools and departments; we are in the James Clark Maxwell Building (JCMB). The school is formed of 3 institutes and 5 research centres including the Higgs centre of Theoretical Physics. We are part of the Institute for Particle and Nuclear Physics, which is itself divided into 3 research group (Nuclear physics, Particle Physics experiments and Particle Physics theory). The ATLAS experiment is part of the Particle Physics experiments group and is formed by 3 academic staff (soon to be 5), 6 researchers and 7 PhD students.

The group, unsurprisingly, focuses its effort on the measurements of the Higgs boson but we now have also activities in the Exotic group and in the Top group of the ATLAS experiment. Our technical contribution is mainly based on the development of the Simulation, with a focus on fast simulation, and on Trigger activities.

Serena Palazzo, our ESR, will mainly work on differential cross-section measurements of top pair production. In line with the group activities, she will also work on fast simulation, but she will give a Machine Learning spin to it. We will use the ever more popular GANs to speed up the ATLAS simulation by several orders of magnitude. Serena will also have a significant period of secondments in companies working with financial data that she will use to improve fraud detection techniques.

Stay tuned for more updates from us!