peterf@2: #
peterf@2: # exper002.py:
peterf@2: #    Wrapper function for GMM model estimation
peterf@2: #
peterf@2: # Author: Peter Foster
peterf@2: # (c) 2015 Peter Foster
peterf@2: #
peterf@2: 
peterf@2: from pylab import *
peterf@2: from gmm import GMM_methods
peterf@2: import pdb
peterf@2: 
peterf@2:     
peterf@2: def exper002_multilabelclassification(Dataset, ConsideredLabels, CrossVal, Settings, numComponentValues=(1,), featureTypeValues=('librosa_mfccs',)):
peterf@2:     np.random.seed(10553)  
peterf@2:         
peterf@2:     EXPER = {}
peterf@2:     for numComponents in numComponentValues:
peterf@2:         for featureType in featureTypeValues:
peterf@2:             parameterTuple = (numComponents,featureType)
peterf@2:             print('Estimating models for numComponents=' + str(numComponents) + 'featureType=' + str(featureType))
peterf@2:             EXPER[parameterTuple] = {}
peterf@2:             #For each track and for each label store presence and absence scores
peterf@2:             EXPER[parameterTuple]['scores'] = {i:{} for i in range(len(Dataset))}    
peterf@2:                                 
peterf@2:             #Iterate through folds
peterf@2:             for trainI, testI in CrossVal:
peterf@2:                 Data = {}
peterf@2:                 Data['trainingData'] = [np.copy(Dataset.ix[i]['features'][featureType]) for i in trainI]
peterf@2:                 NFeaturesPerFile = [features.shape[0] for features in Data['trainingData']]
peterf@2:                 Data['trainingData'] = np.vstack(Data['trainingData'])
peterf@2:                 Data['testDataPerFile'] = [np.copy(Dataset.ix[i]['features'][featureType]) for i in testI]                                                    
peterf@2:                                                 
peterf@2:                 for label in ConsideredLabels:                            
peterf@2:                     Data['trainingTargets'] = [Dataset.ix[i][label] for i in trainI]
peterf@2:                     #Duplicate labels according to number of feature vectors in each file
peterf@2:                     Data['trainingTargets'] = [[target]*N  for (target,N) in zip(Data['trainingTargets'], NFeaturesPerFile)]
peterf@2:                     #Convert data to one large vector / matrix (for GMM code)
peterf@2:                     Data['trainingTargets'] = np.append([], Data['trainingTargets'])                            
peterf@2:                     
peterf@2:                     Result = GMM_methods.GMMTrainAndTest(Data,numComponents=numComponents)
peterf@2:                     
peterf@2:                     for k in Result['fileScores'].keys():
peterf@2:                         assert(np.all(np.isfinite(Result['fileScores'][k])))
peterf@2:                     
peterf@2:                     #Assign scores to output data structure
peterf@2:                     for i,j in zip(testI, range(len(testI))):    
peterf@2:                         EXPER[parameterTuple]['scores'][i][label] = {True:Result['fileScores'][True][j], False:Result['fileScores'][False][j]}
peterf@2:     return EXPER