Unfolding Meeting, 17.11.2010
People: Nikolai Gagunashvili, Tim Adye, Fred Stober, Thomas Peiffer, Christoph Rosemann, Pedro Cipriano, Gero Flucke
Excused: Natalie Milke
1) News
- unfortunately there are no minutes from last meeting
- Natalie, Tim, Pedro and Gero met last week at DESY , discussing about
- a TRUEE interface for RooUnfold,
- Nikolai's blind test proposal
- some RooUnfold internals
- a project for Pedro
2) Developments
- TRUEE interface for RooUnfold
- need some unbinned Response object: Tim will work on that
- Gero (with help from Natalie) will look for the exact way to interface
- seems to be non-trivial...
- Pedro will try to write a configurable standalone executable for RooUnfold
- Thomas mentions about TRUEE
- had problems with TRUEE's pseudo experiment setup
- problem with distributions going to zero still unsolved?
3) Discussion on Blind Test proposal (see also Gero's slides)
- for simple tests tester should indeed kow exact response and acceptance - to generate his own training data from that
- for comparison of algorithms, in the end we need common binning of the output - but we do not have to define it now
- volume of of dispersion ellipsoid and trace of covariance are well known characteristics of experiments from theory of design of experiments
- People:
- Fred Stober can contribute using the SVD method
- Thomas Peiffer has limited ressources...
- Nikolai will define acceptance and resolutions functions as well as generate 'measured data' from them and link all that from the Wiki
- people are encouraged to test and give feedback
4) AOB
- Gero setups doodle to find a meeting December 13.-16.: http://doodle.com/dhdvax4yg5tuge29
- Gero will not be around 17.12.2010 till 31.1.2011
- Nikolai mentions two papers he wrote and will link to the Wiki
- Parametric fitting of data
- Machine leraning approach to unfolding
There are minutes attached to this event.
Show them.
