diff --git a/Code/MonoMutliViewClassifiers/ExecClassif.py b/Code/MonoMutliViewClassifiers/ExecClassif.py
index 97f65faebd42d3471ec93fe7cc7ccb32044c23c4..081e6de06367a3cdd054758bc86b57e91960d738 100644
--- a/Code/MonoMutliViewClassifiers/ExecClassif.py
+++ b/Code/MonoMutliViewClassifiers/ExecClassif.py
@@ -149,7 +149,7 @@ groupFusion = parser.add_argument_group('Fusion arguments')
groupFusion.add_argument('--FU_types', metavar='STRING', action='store',
help='Determine which type of fusion to use, if multiple separate with :',
default='LateFusion:EarlyFusion')
-groupFusion.add_argument('--FU_ealy_methods', metavar='STRING', action='store',
+groupFusion.add_argument('--FU_early_methods', metavar='STRING', action='store',
help='Determine which early fusion method of fusion to use, if multiple separate with :',
default='')
groupFusion.add_argument('--FU_late_methods', metavar='STRING', action='store',
@@ -276,7 +276,7 @@ if "Multiview" in args.CL_type.strip(":"):
else:
benchmark["Multiview"]["Fusion"]["Methods"]["LateFusion"] = args.FU_late_methods.split(":")
if "EarlyFusion" in args.FU_types.split(":"):
- if args.FU_late_methods.split(":") == [""]:
+ if args.FU_early_methods.split(":") == [""]:
benchmark["Multiview"]["Fusion"]["Methods"]["EarlyFusion"] = [name for _, name, isPackage in
pkgutil.iter_modules(["Multiview/Fusion/Methods/EarlyFusionPackage"])
if not isPackage]
@@ -379,7 +379,7 @@ if True:
except:
pass
- if True:
+ try:
if benchmark["Multiview"]["Fusion"]:
if args.CL_algos_monoview !=['']:
monoClassifiers = args.CL_algos_monoview.split(":")
@@ -389,24 +389,13 @@ if True:
elif not gridSearch:
raise ValueError("No config for fusion method given and no gridearch wanted")
else:
- fusionMethodConfigs = [["config"] for method in benchmark["Multiview"]["Fusion"]["Methods"]["LateFusion"]]
- for methodIndex, method in enumerate(benchmark["Multiview"]["Fusion"]["Methods"]["LateFusion"]):
- if args.FU_fixed:
- arguments = {"CL_type": "Fusion",
- "views": views,
- "NB_VIEW": len(views),
- "viewsIndices": viewsIndices,
- "NB_CLASS": len(args.CL_classes.split(":")),
- "LABELS_NAMES": args.CL_classes.split(":"),
- "FusionKWARGS": {"fusionType":"LateFusion", "fusionMethod":method,
- "classifiersNames": args.FU_cl_names.split(":"),
- "classifiersConfigs": monoClassifiersConfigs,
- 'fusionMethodConfig': fusionMethodConfigs[methodIndex]}}
- argumentDictionaries["Multiview"].append(arguments)
- else:
- for combination in itertools.combinations_with_replacement(range(len(monoClassifiers)), NB_VIEW):
- monoClassifiersNamesComb = [monoClassifiers[index] for index in combination]
- monoClassifiersConfigsComb = [monoClassifiersConfigs[index] for index in combination]
+ try:
+ fusionMethodConfigs = [["config"] for method in benchmark["Multiview"]["Fusion"]["Methods"]["LateFusion"]]
+ except:
+ pass
+ try:
+ for methodIndex, method in enumerate(benchmark["Multiview"]["Fusion"]["Methods"]["LateFusion"]):
+ if args.FU_fixed:
arguments = {"CL_type": "Fusion",
"views": views,
"NB_VIEW": len(views),
@@ -414,10 +403,27 @@ if True:
"NB_CLASS": len(args.CL_classes.split(":")),
"LABELS_NAMES": args.CL_classes.split(":"),
"FusionKWARGS": {"fusionType":"LateFusion", "fusionMethod":method,
- "classifiersNames": monoClassifiersNamesComb,
- "classifiersConfigs": monoClassifiersConfigsComb,
+ "classifiersNames": args.FU_cl_names.split(":"),
+ "classifiersConfigs": monoClassifiersConfigs,
'fusionMethodConfig': fusionMethodConfigs[methodIndex]}}
argumentDictionaries["Multiview"].append(arguments)
+ else:
+ for combination in itertools.combinations_with_replacement(range(len(monoClassifiers)), NB_VIEW):
+ monoClassifiersNamesComb = [monoClassifiers[index] for index in combination]
+ monoClassifiersConfigsComb = [monoClassifiersConfigs[index] for index in combination]
+ arguments = {"CL_type": "Fusion",
+ "views": views,
+ "NB_VIEW": len(views),
+ "viewsIndices": viewsIndices,
+ "NB_CLASS": len(args.CL_classes.split(":")),
+ "LABELS_NAMES": args.CL_classes.split(":"),
+ "FusionKWARGS": {"fusionType":"LateFusion", "fusionMethod":method,
+ "classifiersNames": monoClassifiersNamesComb,
+ "classifiersConfigs": monoClassifiersConfigsComb,
+ 'fusionMethodConfig': fusionMethodConfigs[methodIndex]}}
+ argumentDictionaries["Multiview"].append(arguments)
+ except:
+ pass
else:
try:
if benchmark["Multiview"]["Fusion"]["Methods"]["LateFusion"] and benchmark["Multiview"]["Fusion"]["Classifiers"]:
@@ -458,7 +464,7 @@ if True:
argumentDictionaries["Multiview"].append(arguments)
except:
pass
- else:
+ except:
pass
else:
pass
diff --git a/Code/MonoMutliViewClassifiers/MonoviewClassifiers/SCM.py b/Code/MonoMutliViewClassifiers/MonoviewClassifiers/SCM.py
index 8fb459f72ac6ab46b08374a92328488fbcecdf1f..67f0ce4ba5a9404535e46b8f1b934483c8df62c9 100644
--- a/Code/MonoMutliViewClassifiers/MonoviewClassifiers/SCM.py
+++ b/Code/MonoMutliViewClassifiers/MonoviewClassifiers/SCM.py
@@ -100,9 +100,9 @@ def gridSearch(X_train, y_train, nbFolds=4, metric=["accuracy_score", None], nIt
def getConfig(config):
try :
- return "\n\t\t- SCM with max_attributes : "+str(config[0])+", c : "+str(config[1])+", p : "+str(config[2])
+ return "\n\t\t- SCM with max_attributes : "+str(config[0])#+", c : "+str(config[1])+", p : "+str(config[2])
except:
- return "\n\t\t- SCM with max_attributes : "+str(config["0"])+", c : "+str(config["1"])+", p : "+str(config["2"])
+ return "\n\t\t- SCM with max_attributes : "+str(config["0"])#+", c : "+str(config["1"])+", p : "+str(config["2"])
def transformData(dataArray):
diff --git a/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusion.py b/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusion.py
index 650d4e836a3ae892fbba816a876c1a1ed73f6054..a87bab4203ed617da510d39d26024ab4832ca08c 100644
--- a/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusion.py
+++ b/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusion.py
@@ -11,8 +11,8 @@ class EarlyFusionClassifier(object):
if type(monoviewClassifierConfig[0])==dict:
pass
else:
- monoviewClassifierConfig[0] = dict((str(configIndex), config) for configIndex, config in
- enumerate(self.monoviewClassifiersConfig
+ monoviewClassifierConfig[0] = dict((str(configIndex), config[0]) for configIndex, config in
+ enumerate(monoviewClassifierConfig
))
self.monoviewClassifiersConfig = monoviewClassifierConfig[0]
self.monoviewClassifier = None
@@ -24,11 +24,11 @@ class EarlyFusionClassifier(object):
viewsIndices = np.arange(DATASET.get("Metadata").attrs["nbView"])
nbView = len(viewsIndices)
if not usedIndices:
- uesdIndices = range(DATASET.get("Metadata").attrs["datasetLength"])
+ usedIndices = range(DATASET.get("Metadata").attrs["datasetLength"])
if type(weights)== type(None):
weights = np.array([1/nbView for i in range(nbView)])
if sum(weights)!=1:
weights = weights/sum(weights)
- self.monoviewData = np.concatenate([weights[index]*getV(DATASET, viewIndex, usedIndices)
+ self.monoviewData = np.concatenate([getV(DATASET, viewIndex, usedIndices)
for index, viewIndex in enumerate(viewsIndices)], axis=1)
diff --git a/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusionPackage/WeightedLinear.py b/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusionPackage/WeightedLinear.py
index 9927041126f16ed9b09e16edfdb459b9a21e2387..31bdcc3596bbc70a554db1ec90440ae24637468f 100644
--- a/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusionPackage/WeightedLinear.py
+++ b/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/EarlyFusionPackage/WeightedLinear.py
@@ -22,7 +22,7 @@ def gridSearch(DATASET, classificationKWARGS, trainIndices, nIter=30, viewsIndic
if accuracy > bestScore:
bestScore = accuracy
bestConfig = normalizedArray
- return [bestConfig]
+ return [np.array([1.0 for i in range(nbView)])]
class WeightedLinear(EarlyFusionClassifier):
diff --git a/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/LateFusionPackage/SCMForLinear.py b/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/LateFusionPackage/SCMForLinear.py
index dd7b17e7201adb3fcc6afcdb4991c860575f2c59..57888231b715addfed55322a7c20e3811cc8c810 100644
--- a/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/LateFusionPackage/SCMForLinear.py
+++ b/Code/MonoMutliViewClassifiers/Multiview/Fusion/Methods/LateFusionPackage/SCMForLinear.py
@@ -11,6 +11,7 @@ from pyscm.utils import _unpack_binary_bytes_from_ints
from math import ceil
import random
from sklearn.metrics import accuracy_score
+import itertools
def gridSearch(DATASET, classificationKWARGS, trainIndices, nIter=30, viewsIndices=None):
@@ -23,16 +24,17 @@ def gridSearch(DATASET, classificationKWARGS, trainIndices, nIter=30, viewsIndic
max_attributes = random.randint(1, 20)
p = random.random()
model = random.choice(["conjunction", "disjunction"])
+ order = random.randint(1,nbView)
randomWeightsArray = np.random.random_sample(nbView)
normalizedArray = randomWeightsArray/np.sum(randomWeightsArray)
- classificationKWARGS["fusionMethodConfig"][0] = [p, max_attributes, model]
+ classificationKWARGS["fusionMethodConfig"][0] = [p, max_attributes, model, order]
classifier = SCMForLinear(1, **classificationKWARGS)
classifier.fit_hdf5(DATASET, trainIndices, viewsIndices=viewsIndices)
predictedLabels = classifier.predict_hdf5(DATASET, trainIndices, viewsIndices=viewsIndices)
accuracy = accuracy_score(DATASET.get("Labels")[trainIndices], predictedLabels)
if accuracy > bestScore:
bestScore = accuracy
- bestConfig = [p, max_attributes, model]
+ bestConfig = [p, max_attributes, model, order]
return [bestConfig]
@@ -69,7 +71,8 @@ class SCMForLinear(LateFusionClassifier):
for index, viewIndex in enumerate(viewsIndices):
monoviewDecisions[:, index] = self.monoviewClassifiers[index].predict(
getV(DATASET, viewIndex, usedIndices))
- predictedLabels = self.SCMClassifier.predict(monoviewDecisions)
+ features = self.generateInteractions(monoviewDecisions, order=self.order)
+ predictedLabels = self.SCMClassifier.predict(features)
else:
predictedLabels = []
return predictedLabels
@@ -80,16 +83,19 @@ class SCMForLinear(LateFusionClassifier):
p = float(self.config[0])
maxAttributes = int(self.config[1])
modelType = self.config[2]
+ self.order = self.config[3]
nbView = len(viewsIndices)
+
self.SCMClassifier = pyscm.scm.SetCoveringMachine(p=p, max_attributes=maxAttributes, model_type=modelType, verbose=False)
monoViewDecisions = np.zeros((len(usedIndices), nbView), dtype=int)
for index, viewIndex in enumerate(viewsIndices):
monoViewDecisions[:, index] = self.monoviewClassifiers[index].predict(
getV(DATASET, viewIndex, usedIndices))
- featureSequence = [str(featureIndex) for featureIndex in range(monoViewDecisions.shape[1])]
- featureIndexByRule = np.arange(monoViewDecisions.shape[1], dtype=np.uint32)
+ features = self.generateInteractions(monoViewDecisions, order=self.order)
+ featureSequence = [str(featureIndex) for featureIndex in range(features.shape[1])]
+ featureIndexByRule = np.arange(features.shape[1], dtype=np.uint32)
binaryAttributes = LazyBaptisteRuleList(featureSequence, featureIndexByRule)
- packedData = _pack_binary_bytes_to_ints(monoViewDecisions, 64)
+ packedData = _pack_binary_bytes_to_ints(features, 64)
nameb = "temp_scm_fusion"
if not os.path.isfile(nameb):
dsetFile = h5py.File(nameb, "w")
@@ -110,7 +116,7 @@ class SCMForLinear(LateFusionClassifier):
dsetFile.close()
dsetFile = h5py.File(name, "r")
packedDataset = dsetFile.get("temp_scm")
- attributeClassification = BaptisteRuleClassifications(packedDataset, monoViewDecisions.shape[0])
+ attributeClassification = BaptisteRuleClassifications(packedDataset, features.shape[0])
self.SCMClassifier.fit(binaryAttributes, DATASET.get("Labels")[usedIndices], attribute_classifications=attributeClassification)
try:
dsetFile.close()
@@ -118,6 +124,28 @@ class SCMForLinear(LateFusionClassifier):
except:
pass
+ def generateInteractions(self, monoViewDecisions, order=None):
+ if type(order)==type(None):
+ order = monoViewDecisions.shape[1]
+ genratedIntercations = [monoViewDecisions[:,i] for i in range(monoViewDecisions.shape[1])]
+ if order==1:
+ return monoViewDecisions
+ else:
+ for orderIndex in range(order-1):
+ combins = itertools.combinations(range(monoViewDecisions.shape[1]), orderIndex+2)
+ for combin in combins:
+ generatedDecision = monoViewDecisions[:,combin[0]]
+ for index in range(len(combin)-1):
+ if self.config[2]=="disjunction":
+ generatedDecision = np.logical_and(generatedDecision, monoViewDecisions[:,combin[index+1]])
+ else:
+ generatedDecision = np.logical_or(generatedDecision, monoViewDecisions[:,combin[index+1]])
+ genratedIntercations.append(generatedDecision)
+ return np.transpose(np.array(genratedIntercations).astype(np.uint8))
+
+
+
+
def getConfig(self, fusionMethodConfig, monoviewClassifiersNames,monoviewClassifiersConfigs):
configString = "with SCM for linear with max_attributes : "+str(self.config[1])+", p : "+str(self.config[0])+\
" model_type : "+str(self.config[2])+" has chosen "+\
diff --git a/Code/MonoMutliViewClassifiers/Multiview/GetMultiviewDb.py b/Code/MonoMutliViewClassifiers/Multiview/GetMultiviewDb.py
index 6c4a31418c029f92a1054c362765365b11d5a9a7..9d1ec462239af1467ac4907a10fff5b2e29c764a 100644
--- a/Code/MonoMutliViewClassifiers/Multiview/GetMultiviewDb.py
+++ b/Code/MonoMutliViewClassifiers/Multiview/GetMultiviewDb.py
@@ -13,23 +13,27 @@ __author__ = "Baptiste Bauvin"
__status__ = "Prototype" # Production, Development, Prototype
-def getPlausibleDBhdf5(features, pathF, name , NB_CLASS, LABELS_NAME, nbView=4, nbClass=2, datasetLength=100):
- nbFeatures = 300
+def getPlausibleDBhdf5(features, pathF, name , NB_CLASS, LABELS_NAME, nbView=4, nbClass=2, datasetLength=500):
+ nbFeatures = 150
datasetFile = h5py.File(pathF+"Plausible.hdf5", "w")
+ CLASS_LABELS = np.array([0 for i in range(datasetLength/2)]+[1 for i in range(datasetLength/2)])
for viewIndex in range(nbView):
- if viewIndex== 0 :
- viewData = np.array([np.zeros(nbFeatures) for i in range(datasetLength/2)]+[np.ones(nbFeatures) for i in range(datasetLength/2)]).astype(np.uint8)
- viewDset = datasetFile.create_dataset("View"+str(viewIndex), viewData.shape)
- viewDset.attrs["name"] = "View"+str(viewIndex)
- viewDset.attrs["sparse"] = False
- viewDset.attrs["binary"] = True
- else:
- viewData = np.array([np.random.normal(float((viewIndex+1)*10), 0.42, nbFeatures) for i in range(datasetLength/2)]+[np.random.normal(-float((viewIndex+1)*10),0.42,nbFeatures) for j in range(datasetLength/2)])
- viewDset = datasetFile.create_dataset("View"+str(viewIndex), viewData.shape)
- viewDset.attrs["name"] = "View"+str(viewIndex)
- viewDset.attrs["sparse"] = False
- viewDset.attrs["binary"] = False
- CLASS_LABELS = np.array([0 for i in range(datasetLength/2+5)]+[1 for i in range(datasetLength/2-5)])
+ # if viewIndex== 0 :
+ viewData = np.array([np.zeros(nbFeatures) for i in range(datasetLength/2)]+[np.ones(nbFeatures) for i in range(datasetLength/2)]).astype(np.uint8)
+ fakeTrueIndices = np.random.randint(0, datasetLength/2-1, datasetLength/5)
+ fakeFalseIndices = np.random.randint(datasetLength/2, datasetLength, datasetLength/5)
+ viewData[fakeTrueIndices] = np.ones((len(fakeTrueIndices), nbFeatures))
+ viewData[fakeFalseIndices] = np.zeros((len(fakeFalseIndices), nbFeatures))
+ viewDset = datasetFile.create_dataset("View"+str(viewIndex), viewData.shape, data=viewData)
+ viewDset.attrs["name"] = "View"+str(viewIndex)
+ viewDset.attrs["sparse"] = False
+ viewDset.attrs["binary"] = True
+ # else:
+ # viewData = np.array([np.random.normal(float((viewIndex+1)*10), 0.42, nbFeatures) for i in range(datasetLength/2)]+[np.random.normal(-float((viewIndex+1)*10),0.42,nbFeatures) for j in range(datasetLength/2)])
+ # viewDset = datasetFile.create_dataset("View"+str(viewIndex), viewData.shape)
+ # viewDset.attrs["name"] = "View"+str(viewIndex)
+ # viewDset.attrs["sparse"] = False
+ # viewDset.attrs["binary"] = False
labelsDset = datasetFile.create_dataset("Labels", CLASS_LABELS.shape)
labelsDset[...] = CLASS_LABELS
labelsDset.attrs["name"] = "Labels"