config.py

#!/usr/bin/env python3
 
 
 
# @cond SUPPRESS_DOXYGEN
 
"""
 Configuration classes
 
 The classes defined here are used to uniquely define a FEI training.
 Meaning:
  - The global configuration like database prefix, cache mode, monitoring, ... (FeiConfiguration)
  - The reconstructed Particles (Particle)
  - The reconstructed Channels of each particle (DecayChannel)
  - The MVA configuration for each channel (MVAConfiguration)
  - The Cut definitions of each channel (PreCutConfiguration)
  - The Cut definitions of each particle (PostCutConfiguration)
"""
 
import collections
import copy
import itertools
import typing
import basf2
 
# Define classes at top level to make them pickable
# Creates new class via namedtuple, which are like a struct in C
 
FeiConfiguration = collections.namedtuple('FeiConfiguration', 'prefix, cache, monitor, legacy, externTeacher, training')
FeiConfiguration.__new__.__defaults__ = ('FEI_TEST', None, True, None, 'basf2_mva_teacher', False)
FeiConfiguration.__doc__ = "Fei Global Configuration class"
FeiConfiguration.prefix.__doc__ = "The database prefix used for all weight files"
FeiConfiguration.cache.__doc__ = "The stage which is passed as input, it is assumed that all previous stages"\
                                 " do not have to be reconstructed again. Can be either a number or"\
                                 " a filename containing a pickled number or"\
                                 " None in this case the environment variable FEI_STAGE is used."
FeiConfiguration.monitor.__doc__ = "If true, monitor histograms are created"
FeiConfiguration.legacy.__doc__ = "Pass the summary file of a legacy FEI training,"\
                                  " and the algorithm will be able to apply this training."
FeiConfiguration.externTeacher.__doc__ = "Teacher command e.g. basf2_mva_teacher, b2mva-kekcc-cluster-teacher"
FeiConfiguration.training.__doc__ = "If you train the FEI set this to True, otherwise to False"
 
 
MVAConfiguration = collections.namedtuple('MVAConfiguration', 'method, config, variables, target, sPlotVariable')
MVAConfiguration.__new__.__defaults__ = ('FastBDT',
                                         '--nTrees 400  --nCutLevels 10 --nLevels 3 --shrinkage 0.1 --randRatio 0.5',
                                         None, 'isSignal', None)
MVAConfiguration.__doc__ = "Multivariate analysis configuration class."
MVAConfiguration.method.__doc__ = "Method used by MVAInterface."
MVAConfiguration.config.__doc__ = "Method specific configuration string passed to basf2_mva_teacher"
MVAConfiguration.variables.__doc__ = "List of variables from the VariableManager."\
                                     " {} is expanded to one variable per daughter particle."
MVAConfiguration.target.__doc__ = "Target variable from the VariableManager."
MVAConfiguration.sPlotVariable.__doc__ = "Discriminating variable used by sPlot to do data-driven training."
 
 
PreCutConfiguration = collections.namedtuple('PreCutConfiguration', 'userCut, vertexCut, noBackgroundSampling,'
                                             'bestCandidateVariable, bestCandidateCut, bestCandidateMode')
PreCutConfiguration.__new__.__defaults__ = ('', -2, False, None, 0, 'lowest')
PreCutConfiguration.__doc__ = "PreCut configuration class. These cuts is employed before training the mva classifier."
PreCutConfiguration.userCut.__doc__ = "The user cut is passed directly to the ParticleCombiner."\
                                      " Particles which do not pass this cut are immediately discarded."
PreCutConfiguration.vertexCut.__doc__ = "The vertex cut is passed as confidence level to the VertexFitter."
PreCutConfiguration.noBackgroundSampling.__doc__ = "For very pure channels, the background sampling factor is too high" \
                                                   " and the MVA can't be trained. This disables background sampling."
PreCutConfiguration.bestCandidateVariable.__doc__ = "Variable from the VariableManager which is used to rank all candidates."
PreCutConfiguration.bestCandidateMode.__doc__ = "Either lowest or highest."
PreCutConfiguration.bestCandidateCut.__doc__ = "Number of best-candidates to keep after the best-candidate ranking."
 
PostCutConfiguration = collections.namedtuple('PostCutConfiguration', 'value, bestCandidateCut')
PostCutConfiguration.__new__.__defaults__ = (0.0, 0)
PostCutConfiguration.__doc__ = "PostCut configuration class. This cut is employed after the training of the mva classifier."
PostCutConfiguration.value.__doc__ = "Absolute value used to cut on the SignalProbability of each candidate."
PostCutConfiguration.bestCandidateCut.__doc__ = "Number of best-candidates to keep, ranked by SignalProbability."
 
DecayChannel = collections.namedtuple('DecayChannel', 'name, label, decayString, daughters, mvaConfig, preCutConfig, decayModeID')
DecayChannel.__new__.__defaults__ = (None, None, None, None, None, None, None)
DecayChannel.__doc__ = "Decay channel of a Particle."
DecayChannel.name.__doc__ = "str:Name of the channel e.g. :code:`D0:generic_0`"
DecayChannel.label.__doc__ = "Label used to identify the decay channel e.g. for weight files independent of decayModeID"
DecayChannel.decayString.__doc__ = "DecayDescriptor of the channel e.g. D0 -> K+ pi-"
DecayChannel.daughters.__doc__ = "List of daughter particles of the decay channel e.g. [K+, pi-]"
DecayChannel.mvaConfig.__doc__ = "MVAConfiguration object which is used for this channel."
DecayChannel.preCutConfig.__doc__ = "PreCutConfiguration object which is used for this channel."
DecayChannel.decayModeID.__doc__ = "DecayModeID of this channel. Unique ID for each channel of this particle."
 
 
MonitoringVariableBinning = {'mcErrors': ('mcErrors', 513, -0.5, 512.5),
                             'mcParticleStatus': ('mcParticleStatus', 257, -0.5, 256.5),
                             'dM': ('dM', 100, -1.0, 1.0),
                             'dQ': ('dQ', 100, -1.0, 1.0),
                             'abs(dM)': ('abs(dM)', 100, 0.0, 1.0),
                             'abs(dQ)': ('abs(dQ)', 100, 0.0, 1.0),
                             'pionID': ('pionID', 100, 0.0, 1.0),
                             'kaonID': ('kaonID', 100, 0.0, 1.0),
                             'protonID': ('protonID', 100, 0.0, 1.0),
                             'electronID': ('electronID', 100, 0.0, 1.0),
                             'muonID': ('muonID', 100, 0.0, 1.0),
                             'isSignal': ('isSignal', 2, -0.5, 1.5),
                             'isSignalAcceptMissingNeutrino': ('isSignalAcceptMissingNeutrino', 2, -0.5, 1.5),
                             'isPrimarySignal': ('isPrimarySignal', 2, -0.5, 1.5),
                             'chiProb': ('chiProb', 100, 0.0, 1.0),
                             'Mbc': ('Mbc', 100, 5.1, 5.4),
                             'cosThetaBetweenParticleAndNominalB': ('cosThetaBetweenParticleAndNominalB', 100, -10.0, 10.0),
                             'extraInfo(SignalProbability)': ('extraInfo(SignalProbability)', 100, 0.0, 1.0),
                             'extraInfo(decayModeID)': ('extraInfo(decayModeID)', 101, -0.5, 100.5),
                             'extraInfo(uniqueSignal)': ('extraInfo(uniqueSignal)', 2, -0.5, 1.5),
                             'extraInfo(preCut_rank)': ('extraInfo(preCut_rank)', 41, -0.5, 40.5),
                             'extraInfo(postCut_rank)': ('extraInfo(postCut_rank)', 41, -0.5, 40.5),
                             'daughterProductOf(extraInfo(SignalProbability))':
                             ('daughterProductOf(extraInfo(SignalProbability))', 100, 0.0, 1.0),
                             }
 
 
def variables2binnings(variables):
    """
    Convert given variables into a tuples which can be given to VariableToHistogram
    """
    return [MonitoringVariableBinning[v] if v in MonitoringVariableBinning else (v, 100, -10.0, 10.0) for v in variables]
 
 
def variables2binnings_2d(variables):
    """
    Convert given variables into a tuples which can be given to VariableToHistogram
    """
    result = []
    for v1, v2 in variables:
        b1 = MonitoringVariableBinning[v1] if v1 in MonitoringVariableBinning else (v1, 100, -10.0, 10.0)
        b2 = MonitoringVariableBinning[v2] if v2 in MonitoringVariableBinning else (v2, 100, -10.0, 10.0)
        result.append(b1 + b2)
    return result
 
 
def removeJPsiSlash(string: str) -> str:
    """
    Remove the / in the J/psi particle name
    """
    return string.replace('/', '')
 
 
class Particle:
 
    """
    The Particle class is the only class the end-user gets into contact with.
    The user creates an instance of this class for every particle he wants to reconstruct with the FEI algorithm,    and provides MVAConfiguration, PreCutConfiguration and PostCutConfiguration. These can be overwritten per channel.
    """
 
    def __init__(self, identifier: str,
                 mvaConfig: MVAConfiguration,
                 preCutConfig: PreCutConfiguration = PreCutConfiguration(),
                 postCutConfig: PostCutConfiguration = PostCutConfiguration()):
        """
        Creates a Particle without any decay channels. To add decay channels use addChannel method.
            @param identifier is the pdg name of the particle as a string
                   with an optional additional user label separated by ':'
            @param mvaConfig multivariate analysis configuration
            @param preCutConfig intermediate pre cut configuration
            @param postCutConfig post cut configuration
        """
        
        self.identifier = identifier + ':generic' if len(identifier.split(':')) < 2 else identifier
        v = self.identifier.split(':')
        
        self.name = v[0]
        
        self.label = v[1]
        
        self.mvaConfig = mvaConfig
        
        self.channels = []
        
        self.preCutConfig = preCutConfig
        
        self.postCutConfig = postCutConfig
 
    def __eq__(self, a):
        """
        Compares to Particle objects.
        They are equal if their identifier, name, label, all channels, preCutConfig and postCutConfig is equal
        @param a another Particle object
        """
        return (self.identifier == a.identifier and self.name == a.name and self.label == a.label and
                self.channels == a.channels and self.preCutConfig == a.preCutConfig and self.postCutConfig == a.postCutConfig)
 
    def __str__(self):
        """
        Creates a string representation of a Particle object.
        """
        return str((self.identifier, self.channels, self.preCutConfig, self.postCutConfig, self.mvaConfig))
 
    def __hash__(self):
        """
        Creates a hash of a Particle object.
        This is necessary to use this as a key in a dictionary
        """
        return hash((self.identifier, self.channels, self.preCutConfig, self.postCutConfig, self.mvaConfig))
 
    @property
    def daughters(self):
        """ Property returning list of unique daughter particles of all channels """
        return list(frozenset([daughter for channel in self.channels for daughter in channel.daughters]))
 
    def addChannel(self,
                   daughters: typing.Sequence[str],
                   mvaConfig: MVAConfiguration = None,
                   preCutConfig: PreCutConfiguration = None):
        """
        Appends a new decay channel to the Particle object.
            @param daughters is a list of pdg particle names e.g. ['pi+','K-']
            @param mvaConfig multivariate analysis configuration
            @param preCutConfig pre cut configuration object
        """
        # Append generic label to all defined daughters if no label was set yet
        daughters = [d + ':generic' if ':' not in d else d for d in daughters]
        # Use default mvaConfig of this particle if no channel-specific config is given
        mvaConfig = copy.deepcopy(self.mvaConfig if mvaConfig is None else mvaConfig)
        # Use default preCutConfig of this particle if no channel-specific config is given
        preCutConfig = copy.deepcopy(self.preCutConfig if preCutConfig is None else preCutConfig)
        # At the moment all channels must have the same target variable. Why?
        if mvaConfig is not None and mvaConfig.target != self.mvaConfig.target:
            basf2.B2FATAL(
                f'Particle {self.identifier} has common target {self.mvaConfig.target}, while channel '
                f'{" ".join(daughters)} has {mvaConfig.target}. Each particle must have exactly one target!')
        # Replace generic-variables with ordinary variables.
        # All instances of {} are replaced with all combinations of daughter indices
        mvaVars = []
        for v in mvaConfig.variables:
            if v.count('{}') <= len(daughters):
                mvaVars += [v.format(*c) for c in itertools.combinations(list(range(0, len(daughters))), v.count('{}'))]
        mvaConfig = mvaConfig._replace(variables=mvaVars)
        # Add new channel
        decayModeID = len(self.channels)
        self.channels.append(DecayChannel(name=self.identifier + '_' + str(decayModeID),
                                          label=removeJPsiSlash(self.identifier + ' ==> ' + ' '.join(daughters)),
                                          decayString=self.identifier + '_' + str(decayModeID) + ' -> ' + ' '.join(daughters),
                                          daughters=daughters,
                                          mvaConfig=mvaConfig,
                                          preCutConfig=preCutConfig,
                                          decayModeID=decayModeID))
        return self
 
# @endcond