Introduction to neural networks in high energy physics

Artificial neural networks are a well established tool in high energy physics, playing an important role in both online and offline data analysis. Nevertheless they are often perceived as black boxes which perform obscure operations beyond the control of the user, resulting in a skepticism against any results that may be obtained using them. The situation is not helped by common explanations which try to draw analogies between artificial neural networks and the human brain, for the brain is an even more complex black box itself. In this introductory text, I will take a problem-oriented approach to neural network techniques, showing how the fundamental concepts arise naturally from the demand to solve classification tasks which are frequently encountered in high energy physics. Particular attention is devoted to the question how probability theory can be used to control the complexity of neural networks.

[1]  F. Tegenfeldt,et al.  TMVA - Toolkit for multivariate data analysis , 2012 .

[2]  M. Feindt,et al.  The NeuroBayes neural network package , 2006 .

[3]  Eric R. Ziegel,et al.  The Elements of Statistical Learning , 2003, Technometrics.

[4]  Klaus-Robert Müller,et al.  Efficient BackProp , 2012, Neural Networks: Tricks of the Trade.

[5]  Robert B. Ash,et al.  Information Theory , 2020, The SAGE International Encyclopedia of Mass Media and Society.

[6]  David J. C. MacKay,et al.  Information Theory, Inference, and Learning Algorithms , 2004, IEEE Transactions on Information Theory.

[7]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[8]  David J. C. MacKay,et al.  Bayesian Methods for Backpropagation Networks , 1996 .

[9]  Radford M. Neal.  Priors for Infinite Networks , 1996 .

[10]  J. van Leeuwen,et al.  Neural Networks: Tricks of the Trade , 2002, Lecture Notes in Computer Science.

[11]  Radford M. Neal.  Pattern Recognition and Machine Learning , 2007, Technometrics.