.. .. project maintainer: Mike Richman (mike.d.richman@gmail.com) Release Notes ============= main ---- 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 - clean up F523, format has unused arguments (#3299) - general python cleanups (ruff/E713) (#3269) - fix E711 none-comparison (#3267) - Fix a number of issues identified by ruff (#3262) - remove unnecessary `__future__` imports (#3243) - Relax test tolerance on Apple Silicon (#3213) - general python cleanups Sep. 28, 2023 don la dieu (nega at icecube.umd.edu) --------------------------------------------------- IceTray Release v1.8.0 - replace deprecated unittest asserts with modern equivalents (#3168) 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) - General code cleanups Apr. 12, 2023 don la dieu (nega AT icecube.umd.edu) --------------------------------------------------- IceTray Release v1.6.0 - Restore use of BUILD_${PROJECT}=OFF (#3058) Dec. 12, 2022 don la dieu (nega AT icecube.umd.edu) --------------------------------------------------- IceTray Release v1.5.0 - Documentation cleanups Jul. 29, 2022 don la dieu (nega AT icecube.umd.edu) --------------------------------------------------- IceTray Release v1.4.0 - Drop support for Python 2 - Documentation updates - Clean up documentation Feb. 22, 2022 don la dieu (nega AT icecube.umd.edu) --------------------------------------------------- IceTray release v1.3.0 - Test updates Dec. 20, 2019 Alex Olivas (aolivas@umd.edu) ------------------------------------------- Combo Release V00-00-00 November 30, 2017 Kevin Meagher -------------------------------------------------------------------- Release V00-06-04 - add namespaces to ambiguous references - Script cleanup: remove `Finish` and `TrashCan` March 24, 2017 Mike Richman (mike.d.richman@gmail.com) -------------------------------------------------------------------- Release V00-06-03 - Python 3 cleanup -- Kevin Meagher and David Schultz - Added more tests - Refuse to weight by livetime when livetime has not been specified -- Mike Richman May 2, 2016 Alex Olivas -------------------------------------------------------------------- Release V00-06-02 - Modified behavior of Model.score(..., use_purity=True) to follow the SAMME.R prescription - Added access to use_purity=True in PyBDTModule - Enabled DTLearner.use_purity for boosing using purity info (again according to SAMME.R) - Corrected behavior of BDTLearner.add_after_pruner() in C++ - Added pure-python construction of DTModel and BDTModel - Added util.load_sklearn() and resources/scripts/import_from_sklearn.py - More carefully written, and faster, tests * Old tests (try a bunch of calls, check that nothing crashes) are currently disabled -- Mike Richman April 18, 2015 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-06-01 - Ensure that single events are scored with float values expected by C++ -- Mike Richman April 3, 2015 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-06-00 - Add a converter module ConvertDS, so that variables can be pulled from a tray by the same varsfunc-function that they are scored with later on -- Marcel Zoll - Added a #define to prevent numpy 1.7 from issuing warnings about potentially using deprecated APIs -- Claudio Kopper - Correct docs on histlite plotting - Use scipy.special.kolmogorov instead of deprecated kstest - Fix typo so rowintegrals is no longer doubly defined - Add some slicing (Hist2D) and sampling (Hist) functionality to histlite - Add tolerance parameter for bin agreement in Hist.rebin() and use absolute instead of relative error - Use numpy.interp() instead of scipy.interpolate.interp1d() -- Mike Richman - Change code so the run_all_examples script can now be run from anywhere and results will always be in $I3_BUILD/pybdt/resources/examples -- Meike de With - Change classes to structs to fix warnings -- Don la Dieu December 10, 2014 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-05-00 - Add a function to get a subset of dataset and use this in the test - Correct range bug in create_overtrain_check_plot() -- Mike Richman - Fix inline math in docs so it renders with new matplotlib - Let env find bash in the shell scripts -- Don la Dieu September 23, 2014 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-04-02 - Improve logging - Remove -L (use nonlinear cut spacing) from defaults in train_sample_bdt.sh - Catch invalid range early in histlite.Binner() to avoid errors later - For plot_surface(), if cbar is true give the user access to it by returning a dict instead of just the colormesh - Add common pitfalls documentation -- Mike Richman July 11, 2014 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-04-01 - Only build project if numpy is found - Force matplotlib to use Agg backend for compatibility with non-X11 systems - Use nonposy='clip' to simplify and improve logy errorbar handling -- Mike Richman - Check whether matplotlib is available before running `setup_sample_` validator.py -- Meike de With April 22, 2014 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-04-00 - Fix bug where legend does not show line color for labeled objects - Only delete kwargs['label'] if it is present to begin with - Change default DTLearner.num_trees from 50 to 300 - Add new feature: --nonlinear-cuts (attempted cuts are based on statistics rather than linear binning) - Use same binning for all 4 histograms in build_tree - Add zmin kwarg vor plot_surface() -- Mike Richman - Explicitly qualify all shared_ptr as boost::shared_ptr to avoid ambiguity with std::shared_ptr in libc++ -- Claudio Kopper - Explicitly qualify all make_shared as boost::make_shared - Do not run tests dependent on matplotlib and scipy if one of these is not available January 15, 2014 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-03-00 - Add BDTModel.get_subset_bdtmodel_list() for getting an arbitrary subset of the original BDTModel - Add new event_variable_importance() functions, which get the variable importance for the nodes used with a single event, rather than the whole tree or forest - Improved plotting - Change get_values_weights_scores() to get_values_weights() -- Mike Richman - Explicitly qualify all boost::shared_ptrs to make this fully compatible with libc++ -- Claudio Kopper September 30, 2013 Mike Richman (mike.d.richman@gmail.com) -------------------------------------------------------------------- Release V00-02-02 - Bugfix: array overflow when histogramming values very near max value -- Mike Richman - Bugfix: correctly set pointer from DTNodes back to DTModel -- Mike Richman - Bugfix: add quotes around string when using pydot -- Mike Richman August 20, 2013 Meike de With (meike.de.with@desy.de) -------------------------------------------------------------------- Release V00-02-01 - Explicitly use boost::shared_ptr in order not to interfere with std::shared_ptr - Change CMakeLists.txt to make it work with I3_PORTS's cmake -- Claudio Kopper - Make project compatible with Python 3 -- Nathan Whitehorn June 18, 2013, Emanuel Jacobi (emanuel.jacobi@desy.de) -------------------------------------------------------------------- Release V00-02-00 - Add Vine Learner - Make score and pscore calculation optional - In validation account for weights in KS test, check correlation with cuts, ratio plots - Improved plotting - Bugfixes -- Mike Richman April 04, 2013, Emanuel Jacobi (emanuel.jacobi@desy.de) -------------------------------------------------------------------- Release V00-01-01 - Remove bug in the training algorithm which results in using the wrong arrays unless the signal and background datasets contain exactly the same arrays and some other conditions are satisfied. -- Mike Richman February 23, 2013, Emanuel Jacobi (emanuel.jacobi@desy.de) -------------------------------------------------------------------- Release V00-01-00 - Requires python version >= 2.6 -- Mike Richman - remove externals - Include Python.h directly to avoid compilation failures -- Nathan Whitehorn January 17, 2013, Emanuel Jacobi (emanuel.jacobi@desy.de) -------------------------------------------------------------------- - copied from sandbox/richman/pybdt