Reconstruction Algorithm Comparison
 

  » Introduction

  » Monte Carlo sample

  » Processing times



  » Trigger level
      » effective area
      » angular resol. vs energy
      » efficiency
      » energy spectra
      » angular resol. vs declination
      » events vs azimuth
      » fakes
      » summay table

      » comparison 9, 21, 80 lines
      » check retriggering


  » Cut set 1
      » effective area
      » angular resol. vs energy
      » efficiency
      » energy spectra
      » angular resol. vs declination
      » events vs azimuth
      » fakes
      » summay table

      » comparison 9, 21, 80 lines
      » check retriggering


 » Cut set 2
      » effective area
      » angular resol. vs energy
      » efficiency
      » energy spectra
      » angular resol. vs declination
      » events vs azimuth
      » fakes
      » summay table

      » comparison 9, 21, 80 lines
      » check retriggering


 » Cut set 3
      » effective area
      » angular resol. vs energy
      » efficiency
      » energy spectra
      » angular resol. vs declination
      » events vs azimuth
      » fakes
      » summay table

      » comparison 9, 21, 80 lines
      » check retriggering


  »  Event rates  vs  zenith cut
     » corsika events
     » cosmic signal
     » atmospheric neutrinos

  »  Event rates  vs  threshold
 


  »  Preliminary conclusions
 



 Comments to:
   zornoza@icecube.wisc.edu

Introduction

The structure of the page is as follows:

  • Introduction + Monte Carlo + Processing time (this page)
  • Results at trigger level
  • Results using the so-called cutset 1, with |Sdir| < 0.6, Ldir > 150 m and Ndir >= 8
  • Results using the Level 1 cuts (called here cutset 2) from the paper on the IceCube sensitivity (link)
  • Results using the Level 2 cuts (called here cutset 3) from the paper on the IceCube sensitivity (link)
  • Analysis of the trigger rates for different depending on the zenith angle cut.
  • Preliminary conclusions

We show here a comparison among different first-guess reconstruction strategies:

  • Linefit
  • Dipolefit
  • Cluster JAMS*
  • Direct Walk
*Remark: cluster_jams produces a discrete zenith angle distribution, which explains the particular shape of some of the distributions corresponding to this strategy).

We also include results from the muon-llh-reco module fed by each of these first guesses.
The muon-llh-reco result fed by linefit is implicitly assumend when only one strategy is shown.

The comparison is done with different kinds of events:
  • Corsika events
  • Atmospheric neutrinos (Bartol+Naumov)
  • Cosmic neutrinos (E-2)
The magnitudes to be compared are:
  • Effective area
  • Angular resolution
  • Efficiency
  • Fake rate
  • Processing time
We have simulated the 9-line, 21-line and 80-line configurations.

NOTE: There is A LOT of information here, so you can go to the preliminary conclusions for a summary of this (on-going) work.
You can also access the summary tables (trigger, 1, 2, 3 levels) for a quick look at the global results in each case.


MC samples



corsika
nugen_numu
9 lines (at generation)
85 - 1.735e9
 105 - 1.915e6
9 lines (retriggered)
72 - 1.893e9
  79 - 4e5
21 lines (retriggered)
72 - 1.998e9
  80 - 2e5
80 lines
72 - 1.762e9
  78 - 6.14e4
legend: data_set - number of events generated

Processing times

In these tables we sow the processing times per call for each of the



dipolefit
linefit
clusterjams
direct walk
llh (dipole)
llh (line)
llh (cjams)
llh (dw)

9 strings
real time (s)
CPU time (s)
0.000497 0.000518
0.000382 0.000451
0.088036 0.087582
0.006897 0.006747
0.060671 0.060336
0.056960 0.056679
0.062241 0.061593
0.052638 0.052159

21 strings
real time (s) CPU time (s)
0.000485 0.000538
0.000419 0.000538
0.084133 0.083513
0.005437 0.005484
0.054759 0.054444
0.050597 0.050394
0.057006 0.056918
0.042162 0.041792

80 strings
real time (s) CPU time (s)
0.000496 0.000518
0.000380 0.000309
0.071118 0.070558
0.006635 0.006783
0.055136 0.054671
0.054914 0.054731
0.056635 0.056165
0.047390 0.047261