Presentation is loading. Please wait.

Presentation is loading. Please wait.

Criteria for quality assurance and comparison among

Similar presentations


Presentation on theme: "Criteria for quality assurance and comparison among"— Presentation transcript:

1 Criteria for quality assurance and comparison among
Bochum, September 10th , 2013 Criteria for quality assurance and comparison among different pattern recognition programs Gianluigi Boca Pavia University Gianluigi Boca, 10/9/2013

2 General criteria time speed of algorithm;  easy to determine
number of reconstructed tracks; number of ghost tracks; for each reconstructed track : % of the true hits actually present in track; % of true hits missing; % of track hits that are spurious; momentum resolution; for reconstructed track from secondaries : intercepts at z = 0 resolution. easy to define Gianluigi Boca, 10/9/2013

3 The following depend crucially on which MC truth
track is associated to the reconstructed track : number of reconstructed tracks; number of ghost tracks; for each reconstructed track : % of the true hits actually present in track; % of true hits missing; % of track hits that are spurious; momentum resolution; Gianluigi Boca, 10/9/2013

4 Proposal of a criterion to associate a MC truth track
to one (if any) of the reconstructed tracks for tracks coming from (0,0,0) : 1 – consider only tracks coming from interaction vertex and with at least 2 Stt hits; 2 – assign a reconstructed track to a certain MC truth track considering the majority of hits (Stt+Mvd) in common ; 3 – achieve that by constructing the following table : Gianluigi Boca, 10/9/2013

5 Table of the hits (Mvd+Stt) in common between MC
truth tracks and reconstructed tracks Montecarlo Truth tracks MC trk 1 trk 2 trk 3 trk 4 rec trk a 29 rec trk b 4 20 3 rec trk c 1 25 rec trk d 14 2 rec trk e rec trk F 10 rec trk G 26 Reconstructed Tracks

6 Montecarlo Truth tracks
cell with the maximum number in table : assign the reconstructed track A to MC truth track 2. Montecarlo Truth tracks MC trk 1 trk 2 trk 3 trk 4 rec trk a 29 rec trk b 4 20 3 rec trk c 1 25 rec trk d 14 2 rec trk e rec trk F 10 rec trk G 26 Reconstructed Tracks

7 Montecarlo Truth tracks
erase the corresponding row and column from table and obtain a smaller table Montecarlo Truth tracks MC trk 1 trk 2 trk 3 trk 4 rec trk a 29 rec trk b 4 20 3 rec trk c 1 25 rec trk d 14 2 rec trk e rec trk F 10 rec trk G 26 Reconstructed Tracks

8 Montecarlo Truth tracks
MC trk 1 trk 3 trk 4 rec trk b 4 20 3 rec trk c 25 rec trk d 14 2 rec trk e 1 rec trk F 10 rec trk G 26 Reconstructed Tracks

9 Montecarlo Truth tracks
cell with the maximum number in table : assign the reconstructed track G to MC truth track 1. Montecarlo Truth tracks MC trk 1 trk 3 trk 4 rec trk b 4 20 3 rec trk c 25 rec trk d 14 2 rec trk e 1 rec trk F 10 rec trk G 26 Reconstructed Tracks

10 Montecarlo Truth tracks
erase the corresponding row and column from table and obtain a smaller table Montecarlo Truth tracks MC trk 1 trk 3 trk 4 rec trk b 4 20 3 rec trk c 25 rec trk d 14 2 rec trk e 1 rec trk F 10 rec trk G 26 Reconstructed Tracks

11 Montecarlo Truth tracks
MC trk 3 trk 4 rec trk b 20 3 rec trk c 25 rec trk d 2 rec trk e rec trk F 10 Reconstructed Tracks

12 Montecarlo Truth tracks
cell with the maximum number in table : assign the reconstructed track C to MC truth track 4. Montecarlo Truth tracks MC trk 3 trk 4 rec trk b 20 3 rec trk c 25 rec trk d 2 rec trk e rec trk F 10 Reconstructed Tracks

13 Montecarlo Truth tracks
erase the corresponding row and column from table and obtain a smaller table Montecarlo Truth tracks MC trk 3 trk 4 rec trk b 20 3 rec trk c 25 rec trk d 2 rec trk e rec trk F 10 Reconstructed Tracks

14 Montecarlo Truth tracks
MC trk 3 rec trk b 20 rec trk d 2 rec trk e rec trk F 10 Reconstructed Tracks

15 Montecarlo Truth tracks
cell with the maximum number in table : assign the reconstructed track B to MC truth track 3. Montecarlo Truth tracks MC trk 3 rec trk b 20 rec trk d 2 rec trk e rec trk F 10 Reconstructed Tracks

16 Montecarlo Truth tracks
erase the corresponding row and column from table and obtain a smaller table Montecarlo Truth tracks MC trk 3 rec trk b 20 rec trk d 2 rec trk e rec trk F 10 Reconstructed Tracks

17 Montecarlo Truth tracks
erase the corresponding row and column from table and obtain a smaller table Montecarlo Truth tracks rec trk d rec trk e rec trk F Reconstructed Tracks

18 Montecarlo Truth tracks
reconstructed tracks D, E , F are ghost tracks. Montecarlo Truth tracks rec trk d rec trk e rec trk F Reconstructed Tracks

19 When the association criterion is defined, it becomes
straightforward to calculate : number of reconstructed tracks; number of ghost tracks; for each reconstructed track : % of the true hits actually present in track; % of true hits missing; % of track hits that are spurious; momentum resolution; Gianluigi Boca, 10/9/2013

20 A task (written by Lia Lavezzi) already exists doing
essentially the quality check described above : PndTrkQATask.cxx Lia and I have still to check some differencies of results between her task and my proposal. We will sort out that ASAP and then everybody will be able to use it (it is already in svn).


Download ppt "Criteria for quality assurance and comparison among"

Similar presentations


Ads by Google