laura is hosted by Hepforge, IPPP Durham

Opened 10 years ago

Closed 9 years ago

#11 closed task (fixed)

Introduce the ability to float parameters of lineshapes, e.g. masses and widths

Reported by: Thomas Latham Owned by: Thomas Latham
Priority: major Milestone: v3r0
Version: Keywords:
Cc:

Description


Change History (27)

comment:1 Changed 10 years ago by Thomas Latham

Type: defecttask

comment:2 Changed 10 years ago by Thomas Latham

Milestone: Release v3r0
Status: newassigned

comment:3 Changed 10 years ago by Thomas Latham

(In [87]) Commit in work done mainly by Juan (with some help from me) to get something working in terms of floating masses and widths. Works nicely! Currently only in certain cases (RelBW, resonance wider than threshold for special integration treatment). Needs to be generalised (both in terms of lineshapes and also allow all resonance parameters to float not just masses and widths) and made much more efficient. addresses #11

comment:4 Changed 10 years ago by Juan Martin Otalora Goicochea

(In [89]) Commit in changes in ressonances (other than RelBW) to update the initialisation parameters when ressonance's mass or width change. addresses #11

comment:5 Changed 10 years ago by Juan Martin Otalora Goicochea

(In [93]) Change GenFitKpipi to consider correctly new argumets in LauIsobarDinamics::addResonance and Add destructor to the LauFitter class addresses #11

comment:6 Changed 10 years ago by Thomas Latham

(In [116]) Address issue of setting values of resonance parameters for all models. Decided to do away with need to have LauIsobarDynamics know everything. LauIsobarDynamics::addResonance now returns a pointer to LauAbsResonance. Parameters can be changed through LauAbsResonance::setResonanceParameter. LauIsobarDynamics still knows about Blatt-Weisskopf factors - better to only need to set this once. Update GenFit3pi example to demonstrate mechanism. oktotest #34 This should also help with implementing the floating resonance parameters. addresses #11

comment:7 Changed 10 years ago by Juan Martin Otalora Goicochea

(In [118]) Adding the capability to keep the normalization scheme which is store in the new class LauDPPartialIntegralInfo and used when the calcDPNormalisation is called. addresses #11

comment:8 Changed 10 years ago by Thomas Latham

(In [119]) Tidy up code for determining the integration scheme once and using for all reclaculations thereafter, addresses #11. Also fix (in this v3 branch) the bug in the determination of the scheme, addresses #35.

comment:9 Changed 10 years ago by Thomas Latham

(In [126]) Bring the v3r0 branch up to date with the changes to the trunk, addresses #11

comment:10 Changed 10 years ago by Thomas Latham

(In [135]) Large update - now starting to allow other resonance parameters to float. The alpha parameter of the LauBelleNR and LauBelleSymNR lineshapes are working in this version. Will propagate mechanism to other shapes in next update. The mechanism for getting the floating parameters to the fit model is moved from LauResonanceMaker to the the resonances themselves via LauIsobarDynamics. This lays some of the groundwork for tracking which integrals need to be updated. addresses #11

comment:11 Changed 10 years ago by Thomas Latham

(In [136]) Allow the LASS parameters to float and introduce the efficiency saving in the amplitude calculation. Remove R, B, phiR, phiB parameters from complete LASS shape - not needed with the separate shapes available. Add missing method from LauBelleSymNR. Minor fix to doxygen for LauBelleNR. addresses #11 addresses #21

comment:12 Changed 10 years ago by Thomas Latham

(In [137]) Allow parameters to float in separate LASS models. Also introduce more efficient calculation of LASS amplitude. addresses #11 closes #21

comment:13 Changed 10 years ago by Thomas Latham

(In [139]) Add floating of mass and coupling parameters to LauFlatteRes, addresses #11

comment:14 Changed 10 years ago by Thomas Latham

(In [140]) Add floating of parameters to LauKappaRes and LauNRAmplitude, addresses #11

comment:15 Changed 10 years ago by Thomas Latham

(In [141]) Add floating of parameters to LauGounarisSakuraiRes, addresses #11

comment:16 Changed 10 years ago by Daniel Craik

(In [142]) Add floating of parameters to LauSigmaRes and LauDabbaRes, #addresses #11

comment:17 Changed 10 years ago by Thomas Latham

(In [150]) Introduce intelligent recalculation of amplitudes - makes fits with floating resonance parameters more efficient, addresses #11

comment:18 Changed 10 years ago by Thomas Latham

(In [151]) Make resonance parameters second-stage by default. Activate two-stage fit in GenFit3K example. addresses #11

comment:19 Changed 10 years ago by Thomas Latham

(In [155]) Fix problem in toy generation since introducing the smart-integration, addresses #11

comment:20 Changed 10 years ago by Thomas Latham

(In [162]) Fix for the new integration - many thanks to Dan! Few tidy-ups. addresses #11

comment:21 Changed 10 years ago by Thomas Latham

(In [165]) Add new method LauResonanceInfo::createSharedParameterRecord to allow sharing of resonance parameters between components that are not charged conjugates, e.g. LASS_BW and LASS_NR - implemented in LauResonanceMaker and in the example GenFitBelleCPKpipi. addresses #11

comment:22 Changed 10 years ago by Thomas Latham

(In [174]) Fix floating resonance parameters in simultaneous fits, addresses #11

comment:23 Changed 10 years ago by Thomas Latham

Issues still outstanding on this ticket:

  • Blatt--Weisskopf barrier factors need some work to allow them to float
  • Can have large memory usage, O(GBs), if the integration grid is very finely binned (I fear that there is no way around this but at the very least we should try and print warnings if the usage will go over some amount and perhaps try and make the binning more tunable)

Should probably do a full profile and apply any tweaks to improve timing before releasing this.

comment:24 Changed 9 years ago by Thomas Latham

(In [188]) Implement mechanism for floating Blatt-Weisskopf barrier factor radii, addresses #11

comment:25 Changed 9 years ago by Thomas Latham

(In [191]) Add new method to LauResonanceMaker to set whether the radius of a given Blatt-Weisskopf category should be fixed of floated. Modify the methods of LauResonanceMaker to set the radius value and whether it should be floated so that they work before and after the resonances have been created. addresses #11

comment:26 Changed 9 years ago by Thomas Latham

(In [198]) Add methods to make the integration binning more tunable. Print warnings if memory usage is likely to be very large. addresses #11

comment:27 Changed 9 years ago by Thomas Latham

Resolution: fixed
Status: assignedclosed

Profile shows that things are pretty good. One possibility to save a lot of time is to cache all the kinematics information for each grid point and for each event since approx 40-50% of the CPU time is in recalculating all the helicity angles, daughter momenta etc. from the two DP co-ordinates. We'd need to check the memory consumption very carefully but it could be a big saving. Will create a new ticket for this issue and close this one so that v3r0 can be released.

Note: See TracTickets for help on using tickets.