Release Notes¶
main¶
Oct. 24, 2024 don la dieu (nega at icecube.umd.edu)¶
IceTray Release v1.12.0
Cosmetic improvements to stubs (#3494)
June 18, 2024 don la dieu (nega at icecube.umd.edu)¶
IceTray Release v1.11.0
Run mypy in CI (#3480)
mypy cleanup for reconstruction projects (#3418)
Add stubs for even more pybindings (#3413)
update some project dependencies
Feb. 20, 2024 don la dieu (nega at icecube.umd.edu)¶
IceTray Release v1.9.2
Delete mailinglist files (#3357)
Add upgrade rules to ruff (#3316)
Dec. 14, 2023 don la dieu (nega at icecube.umd.edu)¶
IceTray Release v1.9.0
Remove Uber Header (I3.h) (#3151)
Fix #3172 - SPEFit module can’t process user parameters in Debug mode (#3309)
Clean up invalid escape sequences for python3.12 (#3286)
python ruff cleanup fix EXE001 (#3289)
remove unnecessary
__future__imports (#3243)general python cleanups
June 22, 2023 don la dieu (nega at icecube.umd.edu)¶
IceTray Release v1.7.0
Make I3Tray.I3Tray and icecube.icetray.I3Tray the same thing. (#3102)
Remove all references to load_pybindings (#3086)
Dec. 12, 2022 don la dieu (nega AT icecube.umd.edu)¶
IceTray Release v1.5.0
Remove unused import [Tianlu Yuan]
Make the minuit precision a settable argument [Tianlu Yuan]
Switch to a log_warn for iminuit versions < 2 [Tianlu Yuan]
Reintegrate SegmentedSplineReco (#2630) [Jakob van Santen]
Version checking for iminuit [Tianlu Yuan]
Fix up various urls in the documentation [Kevin Meagher]
Update urls in documentation which are redirected [Kevin Meagher]
Use sphinx extension extlinks to shorten/correct links [Kevin Meagher]
Jul. 27, 2022 don la dieu (nega AT icecube.umd.edu)¶
IceTray Release v1.4.0
Drop support for Python 2
Clean up python bindings / modules
Reflect relicensing agreements received from authors of core IceTray modules.
Updates for I3Minuit / iminuit
Updates to C++17
Clean up documentation
March 6, 2020 Alan Coleman (alanc@udel.edu)¶
Release v2.0.0
I3GulliverIPDFPanelandI3GulliverIPDFPanelFactoryremoved from project
March 6, 2020 Alan Coleman (alanc@udel.edu)¶
Combo Release V01-00-00
TMinuit.h/.cxxremoved from LCOV test coverage
December 20, 2019 Alex Olivas (aolivas@umd.edu)¶
Combo Release V00-00-00
December 13, 2019 Alan Coleman (alanc@udel.edu)¶
Release V01-14-03
Expose
I3BasicSeedServiceto PythonChanged the default random number generator from I3SPRNGRandomService to I3GSLRandomService
Added test/dummy minmizer, likelihood, and parameterization which give contrived results to help with development and testing
June 12, 2018 Kevin Meagher¶
Release V01-14-02
Add example of how to use a custom gulliver likelihood service in python using ipdf as an example
PEP 8 compliant tests and examples
December 1, 2017 Kevin Meagher¶
Release V01-14-01
Add support for gradient based minimization algorithms in I3GulliverNLopt
Check for Geometry in
I3BasicSeedService(#2056)Use consistent seed in multinest instead of system clock
move
TMinuit.hinto public directory so stocastics can use itFix integer comparison
Add new example scripts
Clean up existing examples
python3 compatibility
Scripts cleanup: remove
Finish()andTrashCan
Marach 23, 2017 EKB (blaufuss@umd.edu)¶
Release V01-14-00
Fixed bug in TMinuit that caused scripts to crash when
I3GulliverMinuitwas executed in verbose mode.Added fitname parameter to segments so we can stop using the module instance name as the name of the fit.
Added pybindings for
I3GulliverMinuitandI3GulliverMinuit2.Segments now accept parametrization and minimizer services as parameters. (#1898)
docs, python3 cleanups
Serialization from icecube not boost
May 2, 2016 Alex Olivas (olivas@icecube.umd.edu)¶
Release V01-13-00
Always use local, less noisy TMinuit copy even if ROOT is available.
Copy-paste from rdmc into private/test to remove dependency on rdmc.
Deleted I3GulliverRdmcPandel.cxx.
Moved tray segments and minimizers to sub-modules. Those are not automatically imported with lilliput.
New minimizer based on iminuit; implemented in python.
Added I3GulliverNLopt, a wrapper for the NLopt minimization library.
Bug-fix: corrected scaling of tolerance value for Migrad in I3GulliverMinuit2.
Dropped support of ROOT versions less than 5.30 for I3GulliverMinuit2.
April 3, 2015 Meike de With (meike.de.with@desy.de)¶
Release V01-12-00
Remove option to seed from the MCTree.
Apply chain rule correctly for log10 (instead of ln) in I3SimpleParametrization.
Define LBFGSB symbols to prevent scipy from mistaking this for its internal implementation.
– Jakob van Santen
Fixes for new I3MCTree.
– Meike de With
.fits files were moved to I3_TESTDATA, so look for them there.
– Don la Dieu
Recognize I3_TESTDATA.
Replace I3ParticleVectPtr by I3VectorI3ParticlePtr.
– Alex Olivas
Python bindings for simple parametrization.
Small code changes to fix compiler warnings.
Replace most instances of log_{debug,info,notice} by log_trace.
Remove old ROOTCINT things.
– David Boersma
Added documentation.
– Lisa Unger
Add I3GulliverMN, which uses an external version of MultiNest (currently searched for only in system paths, works with homebrew multinest, but will likely not find multinest on linux system with local installation yet). This works with Pegleg to fit the HybridReco hypothesis, but should be used carefully when applied to new problems.
– Martin Leuermann, Claudio Kopper
December 9, 2014 Meike de With (meike.de.with@desy.de)¶
Release V01-11-00
Reduce logging noise in case of empty-missing data.
Fix weird bug where the detectorresponseptr_ would keep its old value if the configured pulsemap does not exist in current frame. So a frame without a pulsemap might still get a fit, namely exactly the same as in the previous frame. However, this only happens if the seed makes the same mistake, which shouldn’t happen.
Write hdf output only if we are not on a buildbot (some of them do not have hdfwriter available).
– David Boersma
Deal with PhotoSplineService API change.
– Jakob van Santen
October 9, 2014 Meike de With (meike.de.with@desy.de)¶
Release V01-10-01
Avoid mixing standalone Minuit2 with ROOT (if USE_ROOT=On and the ROOT install provides Minuit2, use that).
– David Boersma
September 30, 2014 Meike de With (meike.de.with@desy.de)¶
Release V01-10-00
Change ImportError to RuntimeError for dlopen() failures.
Change RuntimeError to ImportError if paraboloid tray segment is not found.
– Jakob van Santen
Replace expandvars by os.path.expandvars.
– Meike de With
Get RST documentation started.
Add python directory.
Add test script for Gulliver Python interface.
Make try_segments.py test script work also if paraboloid is not there.
Remove obsolete cmake_policy lines.
Make sure I3ParaboloidPandelFitter is only attempted if paraboloid exists in the metaproject and it was built with ROOT.
Make tests work with both I3_TESTDATA and I3_PORTS.
Explicitly list test scripts in CMakeLists.txt.
Temporarily disable tests.
No longer use pulse charge in try_segments.py.
– David Boersma
Fix location of testdata.
– Don la Dieu
July 11, 2014 Meike de With (meike.de.with@desy.de)¶
Release V01-09-00
Silence compiler warning about unused variable.
Adding paraboloid segment.
Make noise rate configurable in all segments.
Add disclaimer about functions starting and ending with __.
Change name of some convenience functions that are used by other people so that they no longer have the __, old names are now deprecated and will be removed in a few months.
Suppress some spurious logging noise.
Verify that fits are reproducible.
Print warning at the end if any first guesses were ignored.
– David Boersma
Expand the search for Minuit/Minuit2 to ROOT_LIB_DIR.
– Don la Dieu
April 8, 2014 Meike de With (meike.de.with@desy.de)¶
Release V01-08-00
Cleaning up code.
Fix documentation to reflect code changes.
– Chris Weaver
Add speed as a possible fit parameter.
– David Boersma
January 21, 2014 Meike de With (meike.de.with@desy.de)¶
Release V01-07-00
Make parameterization deal correctly with large values of azimuth and zenith.
– Melanie Day
Add an implementation of the LBFGSB algorithm, for speedy and reliable minimization of numerically-stable likelihoods with gradients.
Remove remaining Fortran I/O leftovers.
Modify the stopping conditions to deal with cases where we know the rough order of magnitude of the function but the dimensions do not all have the same scale.
Wrap scipy minimizers as I3MinimizerBase.
Only check flatness if a variable has returned to its initial value
Remove indexing tricks scale parameter to setulb_().
– Jakob van Santen
Make calls to std::pow unambiguous.
Make project compatible with Python 3.
– Nathan Whitehorn
Ship a private copy of TMinuit in case ROOT is not available, in order to make ROOT-less processing using the older TMinuit SIMPLEX minimizer possible.
– Claudio Kopper
More informative warning messages.
Remove superfluous static global dictionary with lilliput service names and use tray info instead (solves problems in iceprod with multiple trays).
Add options to override the vertex time tweaking (vertex time of improved Linefit is actually good and tweaking on average makes it worse), but default behaviour remains like it was (TFirst).
For iterative fits, if the segment is called with tstype overriding for the input seed(s), make sure pseudorandom iterations are still tweaked with TFirst.
– David Boersma
August 21, 2013 Meike de With (meike.de.with@desy.de)¶
Release V01-06-08
Clean up deprecation warnings related to new dataclasses release.
– David Boersma / Meike de With
Release V01-06-07: June 29, 2013¶
Adding new helper class: parameter with inclusive bounds.
Apparently it can happen with MIGRAD that minimizer->IsValid() returns true even though from its verbose output it is obvious that internally it knows that the result is bogus. Check MinimizerError->IsValid() as well to avoid this.
Release V01-06-06: April 5, 2013¶
Correct a situation where a likelihood could be queried with an invalid angle, but where the result would be transformed back into the valid range before being written to the frame.
Release V01-06-05: February 21, 2013¶
Use I3ServiceBase for I3GSLSimplex and I3SimpleParametrization (to simplify the Factory, avoid code duplication).
Logarithmic E and L will now also use 10-base logs internally (was: natural logs, while 10-based logs were only used in the configuration phase).
Fixed the flatness check: allow for the possibility that during the flatness check a better minimum is found (Claudio).
I3GSLSimplex now uses the I3MinimizerUtils::CheckMinimum to check against flatness.
The code of I3SimpleParametrization(Factory) has now been structured such that it should be easy to add methods for changing individual fitparspecs after construction.
I3SimpleParametrization now has tracing functionality: if you set the “ParticleTrace” option to True then the service will store an I3ParticleVect in the frame with the seed track and all particles that the minimizer tried, for your debugging convenience.
Release V01-06-04: September 11, 2012¶
Use GetName() instead of getting a configuration from the context.
Zero gradients before use, and allow I3HalfSphereParametrization to be used with a fixed direction.
Release V01-06-03: June 25, 2012¶
Allow the proper tests to build if rdmc is not found.
Flatness check should respect parameter bounds (if any).
Make lilliput work with the updated I3OMGeo/I3Orientation.
Release V01-06-02: March 8, 2012¶
Enable MPAll.
Add an option to ignore MIGRADs distance-to-minimum estimate when determining convergence and emit fitted parameters even when the fit fails.
Release V01-06-01: January 11, 2012¶
Preliminary ipdf PEP class that uses photonics-service, so that the usual SPE1st/all and MPE can be done with photospline instead of Pandel.
Support segments.
Use new SetGeometry method.
Dependency cleanups.
Updated tests.
Bugfixes/code optimizations.
Release V01-06-00: August 09, 2011¶
Calculate the Jacobian for the half sphere’s parametrization’s angles correctly.
Introduce variables to store config variables of GSLMultiMin.
Chain rule for I3HalfSphereParametrization.
Release V01-05-01: March 21, 2010¶
Various tweaks and fixes for Minuit2:
Added tests for Minuit2 and made TMinuit tests conditional on USE_ROOT.
Factored flatness check from GulliverMinuit out into a separate function to be used from both kinds of Minuit.
Stopped trusting FunctionMinimum::IsValid() as a measure of convergence. MIGRAD doesn’t seem to lie, but SIMPLEX has a nasty habit of not admitting success. This still needs to be refined.
Release V01-05-00: October 29, 2010¶
Removed the obsolete intlookup feature (integral lookups from a table, for use in MPE/PSA with GConvolute).
ipdf dependency (and hence Pandel availability) is now optional.
Adding simulated annealing minimizer (thanks to Claudio Kopper from KM3NET).
Adding gradient-using GSL minimizers (thanks to Claudio Kopper from KM3NET). We might consider upgrading GSL, as the newest greatest latest algorithm (“vector_bfgs2”) is actually not available in GSL 1.8, what we have in the ports. It seems that the latest stable GSL version is 1.14!
Release V01-04-01: June 04, 2010¶
Fix test. - Fabian Kislat
Release V01-04-00: August 11, 2009¶
Pandel: enable box-convoluted MPE Pandel.
The time calibration uncertainty of a DOM is a systematic uncertainty while PMT jitter is statistical fluctuation from event to event. For SPE this distinction is not relevant, but for MPE it is. The easiest way to take it into account is to assume not a Gaussian smearing but a “box”-like smearing, that is a flat distribution between fixed limits, which is actually quite reasonable for this kind of uncertainty. And as a bonus the box-like smearing is really easy to compute, thanks to the very definition of MPE.
Parametrization:
Implementation work: replace long arg lists with steps and bounds by vectors.
Add “relative” bounds: bounds relative to the initial seed value.
Added extensive unit tests.
Seed service: add VertexMode option, relevant only for ContainedTracks allows to select whether xyz refer to start or stop point (especially important when keeping xyz fixed: keep start point fixed, or stop point?).
Release V01-03-00: October 27, 2008¶
Seed service: polynomial energy guess is NOT converted to energy loss (so if “energy” is actually interpreted by the likelihood function as energy loss (that’s the case for fitting w/ lightsaber tables) then the polynomial you specify should be a fit to the log10(NCh) vs log10(Eloss) distribution, not log10(NCh) vs log10(E).
Seed service: it’s allowed now to configure no seeds at all (that’s useful in case of fg tracks generated by the module itself and then use the seed service for vertex correction etcetera, e.g. paraboloid).
Seed service and pandel service: will not log_fatal anymore in case of completely missing input data (fg tracks, hits/pulses); instead it will count how often it is missing, and report failure statistics at the end (in case of no failures, nothing at all will be reported).
Added “GaussConvolutedFastApproximation” Pandel variety. This requires ipdf release V02-04-06 or trunk (newer than September 2008). It’s “fast” in comparison with the numerically integrating Pandel version, not compared to the (default) “semi-convoluted” version.
Release made for std-processing.
Release V01-02-01: 19 Aug 2008¶
Minor updates, CMakeLists.txt tweak to deal with new cmake version 2.4.6.
Release made for icerec purposes.
Release V01-02-00: June 24, 2008¶
Use the “fast log summer” in Pandel likelihood calcs. FastLogSum uses a numerical trick to compute sums of logs as the log of a product, without risk of running into floating point range limits. This is a bit faster.
Moved all photorec-related stuff to a separate project.
Use the new implementation of Gauss-convoluted Pandel (George, Dipo, Nick).
Added support for cascade (pointlike and directional).
#include the “transverse directions” and “minimum ball” from phys-services rather than from gullliver/utilities (that was a temporary place for it to be implemented).
Release V01-01-00: 3 January, 2008¶
Add new functionality in seed service: generate alternative directions:
Reverse: 1 new track, opposite direction from input track.
Tetrahedron: 3 new tracks, together with input track they are aligned like the faces of a tetrahedron.
Cube: 5 new tracks, together with input track they are aligned like the faces of a cube.
Seed from input track and alternative tracks can have different vertex time corrections. You can choose to use only the alternative tracks, e.g. if you use a log-likelihood fit as input track.
New parametrization service: I3HalfSphereParametrizationService. For now it does exactly what the name says: it parametrizes half of the full direction sphere, with the starting track as the central axis of the hemisphere. In other words: you won’t get further than 90 degrees away from your starting track. In future versions the 90 degrees will be configurable, so that you can stay e.g. within 45 degrees of your starting track. The algebra for that is a bit less simple, will be a bit slower.
Release V01-00-02: 2 November 2007¶
GetName() in parametrization became a const method.
Basic seeder can make “alternatives”: for each good first guess, add e.g. the reverse track (opposite direction), or 3 other tracks such that together with the original track they are aligned like a tetrahedron, or 5 (cubic).
Add I3PhotorecLogLikelihood: like WFLogLikelihood, but then with recopulses.
Change default parametrization: all stepsizes default to 0.
Release V01-00-00: mid September 2007¶
Cleaned up dead code.
Reduce code volume by using I3ServiceBase (get rid of separately defined trivial factories).
Added two options to minuit minimizer instead of using magic numbers in TMinuit verbosity and strategy configuration (code review request).
Further development on I3PhotorecLogLikelihood.
Lots of doxygen and code convention fixes.
TFirst vertex time correction in I3BasicSeedService: don’t do that unconditionally, put a limit on the mean all time residuals, in order to protect against weird early hits (happens often enough to be annoying, completely screws up the reconstruction).
Moved I3DummyFirstGuess module to a different project (reco-toolbox).
Moved parameter file for I3PDFParametrizedPhotonics from resources/data/ to $I3_PORTS/test-data/reconstruction-constants
Implemented many coding convention fixes (as per code review).
Get rid of static global stuff in I3GSLSimplex implementation (the trick to do that is not applicable in I3GulliverMinuit, unfortunately).
Release V00-05-02: August 10, 2007¶
Add new likelihood class: I3PhotorecLogLikelihood (LLH formula very similar as in I3WFLogLikelihood, but then using recopulses).
Fix unit tests.
Start implementing code review suggestions (not finished yet!).
WF new option: choose best WF per DOM per event, instead of always ATWD or always FADC.
Basic seed service: energy estimator based on Nch, instead of constant energy (for fg which do not provide energy)
Release V00-05-01: June 13, 2007¶
WF llh: add WF unit tests
WF llh: split WF saturation option into two: one for PDF and one for the actual WF.
WF llh: various bugfixes in WF llh.
Basic seed service: correct any “Null” shape from FG input.
Release V00-05-00: April 13, 2007¶
WF llh: make default noise prob dependent on WF type.
WF llh: SetEvent more efficient.
WF llh: deal with saturation in PDF.
WF llh: add minimum distance option.
WF llh: reorganize GetLogLikelihood code, split into smaller methods.
WF llh: use per-DOM gains, as from DOMCalibration and DOMStatus (instead of a single fixed gain for all DOMs).
GSL simplex: small fixes.
All: add keywords.
Seed service: big bugfix.
Seed service: code reorganization (following Gulliver interface upgrade), make methods Tweak, FillInTheBlanks and GetCopy.
Photorec pdf: set default finite length.
Parametrized photonics pdf: bugfix.
Parametrized photonics pdf: get rid of minimum energy, more clever Eloss fit.
GSL simplex: allow use
Release V00-04-00: Mon Dec 4 2006¶
Next version of parametrized photonics PDF (infinite cascade) …
Adapt to new I3Calculator::CherenkovCalc interface.
Add wrapper for GSL implementation of the simplex minimizer (still experimental).
Release V00-03-01: Thu Oct 26 2006¶
Release V00-03-00: Thu Oct 26 2006¶
The I4Particle class has been replaced by I3EventHypothesis.
Various improvements in I3WFLogLikelihood.
First version of parametrized photonics PDF (infinite cascade).
