Jump label

Service navigation

Main navigation

You are here:

Main content

Method Development

Machine Learning

Kunst mit Machine Learning

Machine learning and the automatic recognition of inherent patterns play an important role in experimental astroparticle physics, as in many other areas of daily life. Only with the help of suitable learning algorithms can the large amounts of data (big data) be efficiently examined for cosmic messenger particles such as high-energy photons or neutrinos and these can be selected in a suitable way from the overwhelming background of atmospheric muons. For this purpose, the learning algorithms must not only be adequately trained, but their performance must also be validated in an appropriate manner. The work on the automated selection of astrophysical messenger particles is carried out within the framework of the subproject C3 of the Collaborative Research Centre SFB 876. The areas of work in the field of machine learning include the following subpoints:

  • Selection of signal events with ensemble methods
  • Reconstruction of energy and direction through neural networks
  • Spectral reconstruction by learning algorithms
  • Efficiency increase of air shower simulations through machine learning

Visit our Collaborative Research Centre SFB 876 for more information on Deconvolution Algorithms and Inverse Problems in Astroparticle Physics!

You can also read more about end-to-end analysis mit machine learning in this article.

Unfolding Projects developed at E5b

Project specific unfolding algorithms have been developed at E5b, some of which can be viewed openly.

DSEA

This acronym stands for Dortmund Spectrum Estimation Algorithm. Why DSEA is so great.

Funfolding

This is an algorithm that was specifically designed to unfold IceCube data. Key features are "Likelihood-Based unfolding techniques and Decision-Tree based binning". You can find this project on GitHub

TRUEE

TRUEE (Time-dependent Regularized Unfolding for Economics and Engineering problems) is a new software package for the numerical solution of inverse problems (unfolding or deconvolution). The basis for the algorithm is provided by the FORTRAN 77 application RUN (Regularized UNfolding). The unfolding algorithm RUN has been applied in the analysis of particle and astroparticle physics experiments and stood out with notably stable results and reliable uncertainties. Today, besides FORTRAN, the programming language C++ is widely used for analysis software in various research fields. Therefore the C++ unfolding program TRUEE was developed, containing the RUN algorithm and additional extensions to facilitate a comfortable and user-friendly unfolding procedure, due to the powerful C++-based ROOT framework. The results of TRUEE and RUN are identical.

https://www.sciencedirect.com/science/article/pii/S016890021201008X