The Higgs boson discovery at the Large Hadron Collider in 2012 relied on boosted decision trees. Since then, high energy physics (HEP) has applied modern machine learning (ML) techniques to all stages of the data analysis pipeline, from raw data processing to statistical analysis. The unique requirements of HEP data analysis, the availability of high-quality simulators, the complexity of the data structures (which rarely are image-like), the control of uncertainties expected from scientific measurements, and the exabyte-scale datasets require the development of HEP-specific ML techniques. While these developments proceed at full speed along many paths, the nineteen reviews in this book offer a self-contained, pedagogical introduction to ML models' real-life applications in HEP, written by some of the foremost experts in their area.
Written in accessible language without mathematical formulas, this short book provides an overview of the wide and varied applications of artificial intelligence (AI) across the spectrum of physical sciences. Focusing in particular on AI's ability to extract patterns from data, known as machine learning (ML), the book includes a chapter on important machine learning algorithms and their respective applications in physics. It then explores the use of ML across a number of important sub-fields in more detail, ranging from particle, molecular and condensed matter physics, to astrophysics, cosmology and the theory of everything. The book covers such applications as the search for new particles and the detection of gravitational waves from the merging of black holes, and concludes by discussing what the future may hold.
A core principle of physics is knowledge gained from data. Thus, deep learning has instantly entered physics and may become a new paradigm in basic and applied research.This textbook addresses physics students and physicists who want to understand what deep learning actually means, and what is the potential for their own scientific projects. Being familiar with linear algebra and parameter optimization is sufficient to jump-start deep learning. Adopting a pragmatic approach, basic and advanced applications in physics research are described. Also offered are simple hands-on exercises for implementing deep networks for which python code and training data can be downloaded.
Modern analysis of HEP data needs advanced statistical tools to separate signal from background. This is the first book which focuses on machine learning techniques. It will be of interest to almost every high energy physicist, and, due to its coverage, suitable for students.
The discovery in 2012 of the Higgs boson at the Large Hadron Collider (LHC) represents a milestone for the Standard Model (SM) of particle physics. Most of the SM Higgs production and decay rates have been measured at the LHC with increased precision. However, despite its experimental success, the SM is known to be only an effective manifestation of a more fundamental description of nature. The scientific research at the LHC is strongly focused on extending the SM by searching, directly or indirectly, for indications of New Physics. The extensive physics program requires increasingly advanced computational and algorithmic techniques. In the last decades, Machine Learning (ML) methods have made a prominent appearance in the field of particle physics, and promise to address many challenges faced by the LHC. This thesis presents the analysis that led to the observation of the SM Higgs boson decay into pairs of bottom quarks. The analysis exploits the production of a Higgs boson associated with a vector boson whose signatures enable efficient triggering and powerful background reduction. The main strategy to maximise the signal sensitivity is based on a multivariate approach. The analysis is performed on a dataset corresponding to a luminosity of 79.8/fb collected by the ATLAS experiment during Run-2 at a centre-of-mass energy of 13 TeV. An excess of events over the expected background is found with an observed (expected) significance of 4.9 (4.3) standard deviation. A combination with results from other \Hbb searches provides an observed (expected) significance of 5.4 (5.5). The corresponding ratio between the signal yield and the SM expectation is 1.01 +- 0.12 (stat.)+ 0.16-0.15(syst.). The 'observation' analysis was further extended to provide a finer interpretation of the V H(H → bb) signal measurement. The cross sections for the VH production times the H → bb branching ratio have been measured in exclusive regions of phase space. These measurements are used to search for possible deviations from the SM with an effective field theory approach, based on anomalous couplings of the Higgs boson. The results of the cross-section measurements, as well as the constraining of the operators that affect the couplings of the Higgs boson to the vector boson and the bottom quarks, have been documented and discussed in this thesis. This thesis also describes a novel technique for the fast simulation of the forward calorimeter response, based on similarity search methods. Such techniques constitute a branch of ML and include clustering and indexing methods that enable quick and efficient searches for vectors similar to each other. The new simulation approach provides optimal results in terms of detector resolution response and reduces the computational requirements of a standard particles simulation.
A vivid example of the growing need for frontier physics experiments to make use of frontier technology is in the field of Artificial Intelligence (AI) and related themes.By AI we are referring here to the use of computers to deal with complex objects in an environment based on specific rules (Symbolic Manipulation), to assist groups of developers in the design, coding and maintenance of large packages (Software Engineering), to mimic human reasoning and strategy with knowledge bases to make a diagnosis of equipment (Expert Systems) or to implement a model of the brain to solve pattern recognition problems (Neural Networks). These techniques, developed some time ago by AI researchers, are confronted by down-to-earth problems arising in high-energy and nuclear physics. However, similar situations exist in other 'big sciences' such as space research or plasma physics, and common solutions can be applied.The magnitude and complexity of the experiments on the horizon for the end of the century clearly call for the application of AI techniques. Solutions are sought through international collaboration between research and industry.
No basic or applied physics research can be done nowadays without the support of computing systems, ranging from cheap personal computers to large multi-user mainframes. Some research fields like high energy physics would not exist if computers had not been invented. Departing from the more conventional numerical applications, this series of workshops has been initiated to focus on Artificial Intelligence (AI) related developments, such as symbolic manipulation for lengthy and involved algebraic computations, software engineering to assist groups of developers in the design, coding and maintenance of large packages, expert systems to mimic human reasoning and strategy in the diagnosis of equipment or neural networks to implement a model of the brain to solve pattern recognition problems. These techniques, developed some time ago by AI researchers, are confronted by down-to-earth problems arising in high-energy and nuclear physics. All this and more are covered in these proceedings.
The first half deals with the motion of a single particle under the influence of electronic and magnetic fields. The basic language of linear and circular accelerators is developed. The principle of phase stability is introduced along with phase oscillations in linear accelerators and synchrotrons. Presents a treatment of betatron oscillations followed by an excursion into nonlinear dynamics and its application to accelerators. The second half discusses intensity dependent effects, particularly space charge and coherent instabilities. Includes tables of parameters for a selection of accelerators which are used in the numerous problems provided at the end of each chapter.
