Gulliver

Maintainer: Alan Coleman <alanc@udel.edu>

• Release Notes

Gulliver is a Generic Utility for Log-Likelihood-based Reconstructions. This project provides base classes for the components that define a generic log-likelihood based reconstruction in IceCube/AMANDA, and as of 2009 also in KM3NET.

A fitter class I3Gulliver combines these components into a unit that can be used to obtain a single maximum likelihood fit as obtained by running the minimizer starting from a given seed. The “maximum log-likelihood” reconstruction approach in AMANDA/IceCube is explained in more detail in [Ahrens].

This utility was originally created to facilitate the implementation of standard AMANDA reconstruction methods and to stimulate the experimentation with new reconstruction ideas in an easily configurable and transparent way. It does not intend to accommodate really all possible likelihood-based reconstruction algorithms.

Table of contents

• Gulliver components (services)
  • Design Motivations
  • Representation of fit results
• Anatomy of a Gulliver fit
  • Dramatis Personae
  • Act I: Configuration
  • Act II: Getting Ready For The Fit
• How to implement your own minimizer service
  • Overview
  • Language choice
  • Tolerance
  • Maximum number of iterations
  • Convergence
  • Local versus global minimum
  • Gradients
  • Initialization (start of the run)
  • Minimization
  • Result
• How to implement your own log-likelihood service
  • Overview
  • Interface
• How to implement your own parametrization service
  • Overview
  • Physics variables versus fit parameters
  • Gradients
  • API
• How to implement your own seed service
  • Overview
  • API
  • Example

See also

Doxygen documentation, Gulliver modules, Lilliput, FiniteReco, Millipede, Paraboloid.

About the name

Gulliver is a Generic Utility for Log-Likelihood-based Reconstructions. The I, V and E could stand for Ice (or Inspiration), neutrino (\(\nu\) == V) and Everything Else, respectively.

Note

Lemuel Gulliver is the main character in Gulliver’s Travels, a novel written by Jonathan Swift (1726, amended 1735). Originally, the idea was to host implementations of each of the base classes in their own little project, each named after islands and countries in the same novel, such as Lilliput (inhabited by very small people), Brobdingnag (inhabited by giants), Houyhnhnm (governed by horses) and Balnibarbi (mad scientists); but this met with fierce resistance from the IceRec overlord, so I succumbed, and crammed all implemented services into the lilliput project (which was originally only intended for the minimizer services, for obvious reasons).

Have a look at Project Gutenberg.

References

[Ahrens]

AMANDA Collaboration (Ahrens et al.). “Muon Track Reconstruction and Data Selection Techniques in AMANDA” In: Nucl.Instrum.Meth. A524 (2004), pp. 169-194. arXiv:astro-ph/0407044