v0.6.2

New features:

• Reweight existing samples (either generated with MadMiner or standalone MadGraph) through MadMiner.reweight_existing_sample()
• Custom parameter grids / evaluation points in AsymptoticLimits through the new keyword thetas_eval

v0.6.1

New features:

• Dropout support
• Many more activation functions
• Number of workers for data loading can be specified

Bug fixes:

• Fixed crash in DelphesReader and LHEReader when no systematics are used
• Fixed logging error with double parameterized ratio estimation methods

v0.6.0

New features:

• Expanded systematics system. Users now declare systematics with MadMiner.add_systematics(), which in addition to the previous PDF and scale variations also allows normalization uncertainties. When adding samples in the LHEReader and DelphesReader functions, each sample can be linked to an arbitrary subset of systematics, giving the user a lot of flexibility.

Breaking / API changes:

• For processes with systematic uncertainties, the MadMiner file format changed in a not-backward-compatible way. Please do not use files with systematic uncertainties that were generated with MadMiner versions before v0.6.0 with the new code version (MadMiner will crash). Sorry about this.

Bug fixes:

• Fixed wrongly including the central element in the calculation of the PDF uncertainties.

Documentation:

• Updated and expanded tutorial on systematic uncertainties.

Internal changes:

• Some internal changes related to nuisance parameters, including the MadMiner file format.

v0.5.1

New features:

• Automatic shuffling of MadMiner HDF5 files after reading in LHE or Delphes files

Bug fixes:

• Fixed rare crash in AsymptoticLimits

v0.5.0

New features:

• Clean separation between training and validation events: the SampleAugmenter functions have new keywords partition and validation_split. With partition="validation", validation data without potential overlap with the training samples can be generated. In the madminer.ml classes, this can be provided with new keywords like x_val, theta_val when calling train().
• More consistent calculation of the Fisher information covariance: the covariance matrices in mode="score" are now the ensemble covariance, without dividing by sqrt(n_estimators) as before. This is also the default behavior. The old default behavior can be used with mode="modified_score".
• Parameter rescaling also for DoubleParameterizedRatioEstimator and LikelihoodEstimator.
• When continuing the training of a pre-trained model, the parameter and observable rescaling is not overwritten.
• SALLY limits can now be calculated with Ensemble instances of multiple score estimators.

Breaking / API changes:

• The SampleAugmenter functions do no longer accept the keyword switch_train_test_events. Use partition="train" or partition="test" instead.

Bug fixes:

• Fixed bug in the logging output about relative cross-section uncertainties during the Fisher information calculation.
• Fixed MadMiner crashing when calculating adaptive histogram binnings.
• Fixed bug in AsymptoticLimits where only 1 observed event was returned.

Internal changes:

• Abstracted Estimator classes with parameter rescaling into new ConditionalEstimator class.

v0.4.10

New features:

• ParameterizedRatioEstimator now optionally rescales parameters (theta) to zero mean and unit variance during training. Use the keyword rescale_params in ParameterizedRatioEstimator.train().
• Batching of parameter points in the AsymptoticLimits functions now also when using weighted events.

API and breaking changes:

• In MET smearing, the relative term is now multiplied with HT, defined as the scalar sum over the pT of all visible particles (before it was all particles).

Bug fixes:

• Fixed critical bug in the MET calculation in LHEReader, which caused the y component of the MET object to be wrong.
• Fixed crashes when training a likelihood ratio estimator without providing joint score information.

v0.4.9

New features:

• plot_histograms() can now also visualize observed data / Asimov data.
• In 2D parameter spaces, calculating limits with mode="adaptive-sally" now works as described in https://arxiv.org/abs/1805.00020. In higher dimensions, it still just concatenates the scalar product of score and parameter vector with all score components to form a (d+1)-dimensional observable space.

Bug fixes:

• Fixed bug in plot_histograms().

Tutorials and documentation:

• The MadMiner paper is out!
• Updated README and docs, including a new troubleshooting list.
• Cleaned up examples folder.

v0.4.8

New features:

• In AsymptoticLimits, the adaptive histogram binning can now be based on the weights summed over the whole parameter grid instead of just a central point. This is now also the default option.
• New function plot_histograms() in madminer.plotting to visualize the histograms used by AsymptoticLimits.

Bug fixes:

• Substantially improved automatic histogram binning and fixed some numerical issues in AsymptoticLimits functions.

Tutorials and documentation:

• Updated tutorial with new histogram plots.

v0.4.7

New features:

• More observables for LHEReader.add_observable: Users can use "p_truth" to access particles before smearing, and (at least with XML parsing) there are new global observables "alpha_qcd", "alpha_qed", "scale". LHEReader.add_observable_from_function() now accepts functions that take unsmeared particles as first argument.

Bug fixes:

• Fixed bug in sample_train_ratio() with return_individual_n_effective=True
• Fixed bug in DelphesReader when no events survive cuts

Tutorials and documentation:

• Removed outdated Docker link from docs
• Changed morphing basis in particle physics tutorial to work around a weird bug inn the MG-Pythia interface, see #371

Internal changes:

• Refactored LHE parsing. LHE files are now not read into memory all at once, but sequentially.

v0.4.6

New features:

• AsymptoticLimits now supports the SALLINO method, estimating the likelihood with one-dimensional histograms of the scalar product of theta and the estimated score.
• Improved default histogram binning in AsymptoticLimits and added more binning options, including fully manual specification of the binning.
• Histograms now calculate a rough approximate of statistical uncertainties in each bin and give out a warning if it’s large. (At DEBUG logging level they’ll also print the uncertainties always, and Histogram.histo_uncertainties lets the user access the uncertainties.)

Breaking / API changes:

• The AsymptoticLimits functions expected_limits() and observed_limits() now return (theta_grid, p_values, i_ml, llr_kin, log_likelihood_rate, histos). histos is a list of histogram classes, the tutorial shows how they allow us to plot the histograms. The returns keyword to these functions is removed. The keywords theta_ranges and resolutions were renamed to grid_ranges and grid_resolutions.
• Changed some ML default settings: less hidden layers, smaller batch size.
• Changed function names in the FisherInformation class (the old names are still available as aliases for now, but deprecated).

Bug fixes:

• Various small bug fixes.

Tutorials and documentation:

• AsymptoticLimits is finally properly documented.
• All incomplete user-facing docstrings were updated.

Internal changes:

• Refactored histogram class.
• AsymptoticLimits is now much more memory efficient.