Statistical Learning in High Energy and Astrophysics
b"This thesis studies the performance of statistical learning methods\\nin high energy and astrophysics where they have become a standard tool in physics analysis.\\nThey are used to perform complex classification or regression by intelligent pattern recognition.\\nThis kind of artificial intelligence is achieved by the principle ``learning from examples'':\\nThe examples describe the relationship between detector events and their classification.\\n\\nThe application of statistical learning methods is either motivated by the lack of\\nknowledge about this relationship or by tight time restrictions. In the\\nfirst case learning from examples is the only possibility since no theory is available\\nwhich would allow to build an algorithm in the classical way. In the second case\\na classical algorithm exists but is too slow to cope with the time restrictions.\\nIt is therefore replaced by a pattern recognition machine which implements a\\nfast statistical learning method.\\nBut even in applications where some kind of classical algorithm had done a good job,\\nstatistical learning methods convinced by their remarkable performance.\\n\\nThis thesis gives an introduction to statistical learning methods and how they\\nare applied correctly in physics analysis. Their flexibility and high performance will\\nbe discussed by showing intriguing results from high energy and astrophysics.\\nThese include the development of highly efficient triggers, powerful purification of\\nevent samples and exact reconstruction of hidden event parameters.\\n\\nThe presented studies also show typical problems in the application of\\nstatistical learning methods. They should be only second choice in all cases\\nwhere an algorithm based on prior knowledge exists.\\nSome examples in physics analyses are found where these methods are not used\\nin the right way leading either to wrong predictions or bad performance. Physicists also\\noften hesitate to profit from these methods because they fear that\\nstatistical learning methods cannot be controlled in a physically correct way.\\nBesides there are many different statistical learning methods to choose from\\nand all the different methods have\\ntheir advantages and disadvantages -- compared to each other and to classical\\nalgorithms.\\n\\nBy discussing several examples from high energy and astrophysics experiments\\nthe principles, advantages and weaknesses of all popular statistical learning methods\\nwill be analysed. A focus will be put on neural networks as they form some kind of standard\\namong different learning methods in physics analysis."