M. Stoye

Jet flavour classification is of paramount importance for a broad range of applications in modern-day high-energy-physics experiments, particularly at the LHC. In this paper we propose a novel architecture for this task that exploits modern deep learning techniques. This new model, called DeepJet, overcomes the limitations in input size that affected previous approaches. As a result, the heavy flavour classification performance improves, and the model is extended to also perform quark-gluon tagging.

We extend recent work (Brehmer, et. al., 2018) that use neural networks as surrogate models for likelihood-free inference. As in the previous work, we exploit the fact that the joint likelihood ratio and joint score, conditioned on both observed and latent variables, can often be extracted from an implicit generative model or simulator to augment the training data for these surrogate models. We show how this augmented training data can be used to provide a new cross-entropy estimator, which provides improved sample efficiency compared to previous loss functions exploiting this augmented training data.

Deep learning has led to several breakthroughs outside the field of high energy physics, yet in jet reconstruction for the CMS experiment at the CERN LHC it has not been used so far. This report shows results of applying deep learning strategies to jet reconstruction at the stage of identifying the original parton association of the jet (jet tagging), which is crucial for physics analyses at the LHC experiments. We introduce a custom deep neural network architecture for jet tagging. We compare the performance of this novel method with the other established approaches at CMS and show that the proposed strategy provides a significant improvement. The strategy provides the first multi-class classifier, instead of the few binary classifiers that previously were used, and thus yields more information and in a more convenient way. The performance results obtained with simulation imply a significant improvement for a large number of important physics analysis at the CMS experiment.

The positions of the silicon modules of the CMS tracker will be known to O(100 μm) from survey measurements, mounting precision and the hardware alignment system. However, in order to fully exploit the capabilities of the tracker, these positions need to be known to a precision of a few μm. Only a track-based alignment procedure can reach this required precision. Such an alignment procedure is a major challenge given that about 50000 geometry constants need to be measured. Making use of the novel χ2 minimization program Millepede II an alignment strategy has been developed in which all detector components are aligned simultaneously and all correlations between their position parameters taken into account. Different simulated data, such as Z0 decays and muons originated in air showers were used for the study. Additionally information about the mechanical structure of the tracker, and initial position uncertainties have been used as input for the alignment procedure. A proof of concept of this alignment strategy is demonstrated using simulated data.

The positions of the silicon modules of the CMS tracker will be known to 0(100 μm) from survey measurements, mounting precision and the hardware alignment system. However, in order to fully exploit the capabilities of the tracker, these positions need to be known to a precision of a few μm. Only a track-based alignment procedure can reach this required precision. Such an alignment procedure is a major challenge given that about 50.000 geometry constants need to be measured. Making use of the novel χ2 minimization program Millepede II an alignment strategy has been developed in which all detector components are aligned simultaneously and all correlations between their position parameters taken into account. Tracks from different sources such as Z0 decays and cosmic ray muons, plus information about the mechanical structure of the tracker, and initial position uncertainties have been used as input for the alignment procedure. A proof of concept of this alignment strategy is demonstrated using simulated data.

The Alignment of the CMS tracker will require to determine about 10 5 alignment parameters. The MILLEPEDE program, a linear least-squares algorithm, is a promising c andidate for this task, having been used successfully for alignment in several experiments. However, due to the inversion of a large matrix of linear equations, MILLEPEDE in its original form was limited to problems with about 10 4 parameters. A new version of the program, MILLEPEDE II, provides an iterative method to determine the solution of the matrix, which should work for systems with 10 5 parameters, if the matrix is sparse. This method is tested within the CMS object oriented reconstruction framework (ORCA). Its precision and CPU needs are studied and compared to the inversion method, using alignment scenarios of the CMS tracker with currently up to 12000 parameters.

We extend recent work (Brehmer, et. al., 2018) that use neural networks as surrogate models for likelihood-free inference. As in the previous work, we exploit the fact that the joint likelihood ratio and joint score, conditioned on both observed and latent variables, can often be extracted from an implicit generative model or simulator to augment the training data for these surrogate models. We show how this augmented training data can be used to provide a new cross-entropy estimator, which provides improved sample efficiency compared to previous loss functions exploiting this augmented training data.

Purpose. Non-syndromic hereditary ocular diseases can lead to severe functional limitations of vision at all ages. Therefore, molecular genetic research into the causes of hereditary ocular diseases is of paramount importance. Due to the increasingly optimised pathophysiologically adapted treatments and gene therapy options, a rapid determination of the genetic cause of hereditary ocular diseases is neces- sary, also to increase the chances of treatment at early stages of the disease. In this context, the authors retrospectively analysed the database of the “rare diseases” consultation hour of the PraxisKlinik Augenärzte Halle, which has been in existence since 2016, for patients with a suspected hereditary ophthalmological disease after molecular genetic diagnostics, counselling and therapy options. Material and Methods. 367 index patients were treated dur- ing the period under consideration. In 221 (60.2 %) of them, the molecular genetic cause of the underlying disease could be found. In addition to a large group of patients with cor- neal dystrophies (304 patients, detection rate 61.2 %), we treated 23 index patients with hereditary retinal diseases (detection rate 69.6 %), 28 index patients with complex mal- formations (detection rate 50 %) and 15 index patients with optic atrophies (detection rate 33.4 %). In addition to the ophthalmological-clinical examination and diagnostics, ge- nealogical analysis and molecular genetic testing were always performed. Results. On the basis of 3 case studies with a molecularly ge- netically proven hereditary retinal disease (X-linked recessive inherited juvenile retinoschisis, hemizygous missense muta- tion in the RS1 gene; enhanced S-Cone syndrome, compound heterozygous mutation in the RS2E3 gene; oculocutaneous albinism, homozygous mutation in the TYR gene), the inher- itance, the clinical picture, the course and the therapeutic and rehabilitative options were explained. Conclusion. Ophthalmogenetic counselling and diagnostics in cases of suspected hereditary eye diseases are crucial for a prompt and a sufficient possible therapy and optical reha- bilitation in affected families. Thorough history analysis and diagnostics can optimise the detection rate and provide rapid and sufficient patient care. Keywords hereditary retinal disease, homozygous mutation, ophthal- mogenetics, juvenile retinischisis, oculocutaneous albinism

The present status of searches for SUSY in multi‐jet events with missing transverse energy at CMS will be reviewed with particular focus on robust analysis techniques that are suited for early physics data. The importance of topological variables for controlling the overwhelming background from QCD events and data‐driven methods for the estimation of the remaining backgrounds will be laid out. Estimates of the search reach for the early data taking period will be presented.

The CMS experiment at the Large Hadron Collider at CERN is in the last phase of its construction. The harsh radiation environment at LHC will put strong demands in radiation hardness to the innermost parts of the detector. To assess the performance of irradiated silicon microstrip detector modules, a testbeam was conducted at the Testbeam 22 facility of the DESY research center. The primary objective was the signal-to-noise measurement of fully irradiated CMS Tracker modules to ensure their functionality up to 10 years of LHC operation. The paper briefly summarises the basic setup at the facility and the hardware and software used to collect and analyse the data. Some interesting subsidiary results are shown, which confirm the expected behaviour of the detector with respect to the signal-to-noise performance over the active detector area and for different electron energies. The main focus of the paper are the results of the signal-to-noise over reverse bias voltage measurements for CMS Tracker Modules which were exposed to different radiation doses.

This dataset is associated with the LPCC IML (Lhc Physics Center at Cern Inter-experimental Machine Learning) working group. It was produced for the second IML annual workshop (April 2018). This dataset is part of a machine learning "challenge" on jet mass regression at future circular collider (FCC) conditions. Further details can be found on the challenge page, here: https://gitlab.cern.ch/IML-WG/IML_challenge_2018/wikis/home

This dataset is associated with the LPCC IML (Lhc Physics Center at Cern Inter-experimental Machine Learning) working group. It was produced for the second IML annual workshop (April 2018). This dataset is part of a machine learning "challenge" on jet mass regression at future circular collider (FCC) conditions. Further details can be found on the challenge page, here: https://gitlab.cern.ch/IML-WG/IML_challenge_2018/wikis/home