print 'Clustering 3 classes : \n'
#nombre de clusters
nb_cluster = 3
#N.B : difference qu'on souhaite evaluer sur les phonemes : separation consonnes/voyelles = 0, voisee/non-voisee = 1, categories = 2
#KMEANS non initialise 3 classes
clus = MiniBatchKMeans(n_clusters = nb_cluster, init='k-means++', batch_size=700,
n_init=10, max_no_improvement=10, verbose=0)
f = open(fichier, "a")
f.write("KMEANS non initialise 3 clusters\n")
f.close()
clus.fit(X)
labels = clus.labels_
pourcentage(Y , nb_cluster, labels , dict_path , 0, fichier)
pourcentage(Y , nb_cluster, labels , dict_path , 1, fichier)
pourcentage(Y , nb_cluster, labels , dict_path , 2, fichier)
#KMEANS initialise 3 classes
sous = initialisation_centres(nb_cluster, X)
clus = MiniBatchKMeans(n_clusters = nb_cluster, init=sous, batch_size=700,
n_init=10, max_no_improvement=10, verbose=0)
f = open(fichier, "a")
f.write("KMEANS initialise 3 clusters\n")
f.close()
clus.fit(X)
labels = clus.labels_
pourcentage(Y , nb_cluster, labels , dict_path , 0, fichier)
pourcentage(Y , nb_cluster, labels , dict_path , 1, fichier)
pourcentage(Y , nb_cluster, labels , dict_path , 2, fichier)
print "Clustering miniBatch K-means non supervise : \n"
# nombre de clusters
nb_cluster = 3
# N.B : difference qu'on souhaite evaluer sur les phonemes : separation consonnes/voyelles = 0, voisee/non-voisee = 1, categories = 2
# KMEANS non initialise 3 classes
clus = MiniBatchKMeans(
n_clusters=nb_cluster, init="k-means++", batch_size=700, n_init=10, max_no_improvement=10, verbose=0
)
f = open(fichier, "a")
f.write("KMEANS non initialise 3 clusters\n")
f.close()
clus.fit(X)
labels = clus.labels_
pourcentage(Y, nb_cluster, labels, path_dict, 0, fichier)
pourcentage(Y, nb_cluster, labels, path_dict, 1, fichier)
# KMEANS initialise 3 classes
print "Clustering miniBatch k-means supervise : \n"
sous = initialisation_centres(nb_cluster, X)
clus = MiniBatchKMeans(n_clusters=nb_cluster, init=sous, batch_size=700, n_init=10, max_no_improvement=10, verbose=0)
f = open(fichier, "a")
f.write("KMEANS initialise 3 clusters\n")
f.close()
clus.fit(X)
labels = clus.labels_
pourcentage(Y, nb_cluster, labels, path_dict, 0, fichier)
pourcentage(Y, nb_cluster, labels, path_dict, 1, fichier)
# Agglomerative clustering 3 classes
#N.B : difference qu'on souhaite evaluer sur les phonemes : separation consonnes/voyelles = 0, voisee/non-voisee = 1, categories = 2
#KMEANS non initialise 3 classes
clus = MiniBatchKMeans(n_clusters=nb_cluster,
init='k-means++',
batch_size=700,
n_init=10,
max_no_improvement=10,
verbose=0)
f = open(fichier, "a")
f.write("KMEANS non initialise 3 clusters\n")
f.close()
clus.fit(X)
labels = clus.labels_
pourcentage(Y, nb_cluster, labels, path_dict, 0, fichier)
pourcentage(Y, nb_cluster, labels, path_dict, 1, fichier)
#KMEANS initialise 3 classes
print 'Clustering miniBatch k-means supervise : \n'
sous = initialisation_centres(nb_cluster, X)
clus = MiniBatchKMeans(n_clusters=nb_cluster,
init=sous,
batch_size=700,
n_init=10,
max_no_improvement=10,
verbose=0)
f = open(fichier, "a")
f.write("KMEANS initialise 3 clusters\n")
f.close()
clus.fit(X)
# Read mat file and align file. filename = '../data/Bref80_L4M01.mat' alignfile = '../data/Bref80_L4M01.aligned' fbank = sio.loadmat(filename)['d1'] csv = "../resultats/resultatsClustering/matlabFbank.csv" classementPath = "../data/classement" hop_span = 0.01 Y = utiles.getY(fbank, alignfile, hop_span) #Kmeans without initialization 3 classes (consonnes et voyelles) n_clusters = 3 kmeans = KMeans(init='k-means++', n_clusters=n_clusters, n_init=10) kmeans.fit(fbank) centroids = kmeans.cluster_centers_ labels = kmeans.labels_ analyse.pourcentage(Y, n_clusters, labels, classementPath, 0, csv) #Kmeans without initialization 3 classes n_clusters = 3 kmeans = KMeans(init='k-means++', n_clusters=n_clusters, n_init=10) kmeans.fit(fbank) centroids = kmeans.cluster_centers_ labels = kmeans.labels_ analyse.pourcentage(Y, n_clusters, labels, classementPath, 1, csv) #Kmeans withous initialzation 6 classes n_clusters = 6 kmeans = KMeans(init='k-means++', n_clusters=n_clusters, n_init=10) kmeans.fit(fbank) centroids = kmeans.cluster_centers_
# Read mat file and align file. filename = '../data/Bref80_L4M01.mat' alignfile = '../data/Bref80_L4M01.aligned' fbank = sio.loadmat(filename)['d1'] csv = "../resultats/resultatsClustering/matlabFbank.csv" classementPath = "../data/classement" hop_span = 0.01 Y = utiles.getY(fbank,alignfile,hop_span) #Kmeans without initialization 3 classes (consonnes et voyelles) n_clusters = 3 kmeans = KMeans(init='k-means++', n_clusters=n_clusters, n_init=10) kmeans.fit(fbank) centroids = kmeans.cluster_centers_ labels = kmeans.labels_ analyse.pourcentage(Y,n_clusters,labels,classementPath,0,csv) #Kmeans without initialization 3 classes n_clusters = 3 kmeans = KMeans(init='k-means++', n_clusters=n_clusters, n_init=10) kmeans.fit(fbank) centroids = kmeans.cluster_centers_ labels = kmeans.labels_ analyse.pourcentage(Y,n_clusters,labels,classementPath,1,csv) #Kmeans withous initialzation 6 classes n_clusters = 6 kmeans = KMeans(init='k-means++', n_clusters=n_clusters, n_init=10) kmeans.fit(fbank) centroids = kmeans.cluster_centers_