def _train_clustering(point_set, distance_name, k):
labels = np.array([0]*len(point_set))
features = np.zeros((2, len(point_set)))
for i in xrange(len(point_set)):
features[0, i] = point_set[i]['x']
features[1, i] = point_set[i]['y']
labels[i] = point_set[i]['label']
lab = sg.BinaryLabels(labels)
train = sg.RealFeatures(features)
if distance_name == "EuclideanDistance":
distance = sg.EuclideanDistance(train, train)
elif distance_name == "ManhattanMetric":
distance = sg.ManhattanMetric(train, train)
elif distance_name == "JensenMetric":
distance = sg.JensenMetric(train, train)
else:
raise TypeError
kmeans = sg.KMeans(k, distance)
kmeans.train()
return kmeans
def shogunProcess(clustersNumber, dataLessTarget, datasetName, runinfo = None, initialClusters = None):
import shogun
outputFile = datasetOutFile(datasetName, SHOGUN_ALGO, runinfo=runinfo)
if os.path.exists(outputFile):
print("shogun skipped")
return
train_features = shogun.RealFeatures(dataLessTarget.values.astype("float64").transpose())
# distance metric over feature matrix - Euclidean distance
distance = shogun.EuclideanDistance(train_features, train_features)
hierarchical = shogun.Hierarchical(clustersNumber, distance)
#TODO Makes the pyhon process dies!!!???!!!
d = hierarchical.get_merge_distances()
cp = hierarchical.get_cluster_pairs()
with open(outputFile, 'w') as csvfile:
filewriter = csv.writer(csvfile, quoting=csv.QUOTE_MINIMAL)
for index, row in dataLessTarget.iterrows():
filewriter.writerow([index, result[index].item(0)])
def run_kmeans_base(self,
nb_clusters,
src_file,
data_without_target,
dataset_name,
run_number,
config_function,
run_info=None,
nb_iterations=None):
self._init()
output_file, centroids_file = self._prepare_files(
dataset_name, run_info, True)
train_features = shogun.RealFeatures(
data_without_target.values.astype("float64").transpose())
# distance metric over feature matrix - Euclidean distance
distance = shogun.EuclideanDistance(train_features, train_features)
# KMeans object created
kmeans = shogun.KMeans(nb_clusters, distance)
if config_function is not None:
config_function(kmeans)
if nb_iterations is not None:
kmeans.set_max_iter(nb_iterations)
centers, result = Shogun._kmeans_process(kmeans)
ClusteringToolkit._save_clustering(
Shogun._clustering_to_list(data_without_target, result),
output_file)
ClusteringToolkit._save_centroids(Shogun._centroids_to_list(centers),
centroids_file)
return output_file, {"centroids": centroids_file}
def shogunProcess(clustersNumber,
dataLessTarget,
datasetName,
runinfo=None,
initialClusters=None):
import shogun
outputFile = datasetOutFile(datasetName, SHOGUN_ALGO, runinfo=runinfo)
clustersOutputFile = datasetOutFile(datasetName,
centroidFor(SHOGUN_ALGO),
runinfo=runinfo)
if os.path.exists(outputFile) and os.path.exists(clustersOutputFile):
print("shogun skipped")
return
train_features = shogun.RealFeatures(
dataLessTarget.values.astype("float64").transpose())
# distance metric over feature matrix - Euclidean distance
distance = shogun.EuclideanDistance(train_features, train_features)
# KMeans object created
kmeans = shogun.KMeans(clustersNumber, distance)
if initialClusters is None:
# set KMeans++ flag
kmeans.set_use_kmeanspp(True)
else:
# set new initial centers
kmeans.set_initial_centers(
initialClusters.astype("float64").transpose())
# KMeans training
kmeans.train()
# cluster centers
centers = kmeans.get_cluster_centers()
# Labels for data points
result = kmeans.apply()
with open(outputFile, 'w') as csvfile:
filewriter = csv.writer(csvfile, quoting=csv.QUOTE_MINIMAL)
for index, row in dataLessTarget.iterrows():
filewriter.writerow([index, result[index].item(0)])
with open(clustersOutputFile, 'w') as clusterFile:
filewriter = csv.writer(clusterFile, quoting=csv.QUOTE_MINIMAL)
for row in centers.transpose():
filewriter.writerow(row.tolist())
def run_hierarchical(self,
nb_clusters,
src_file,
data_without_target,
dataset_name,
run_number,
run_info=None):
output_file, = self._prepare_files(dataset_name, run_info, False)
train_features = shogun.RealFeatures(
data_without_target.values.astype("float64").transpose())
# distance metric over feature matrix - Euclidean distance
distance = shogun.EuclideanDistance(train_features, train_features)
hierarchical = shogun.Hierarchical(nb_clusters, distance)
# load data feature_matrix = data.swissroll() # create features instance features = sg.RealFeatures(feature_matrix) # create Diffusion Maps converter instance converter = sg.DiffusionMaps() # set target dimensionality converter.set_target_dim(2) # set number of time-steps converter.set_t(2) # set width of gaussian kernel converter.set_width(10.0) # create euclidean distance instance distance = sg.EuclideanDistance() # enable converter instance to use created distance instance converter.set_distance(distance) # compute embedding with Diffusion Maps method embedding = converter.embed(features) # compute custom distance matrix distance_matrix = np.exp(-np.dot(feature_matrix.T,feature_matrix)) # create Custom Kernel instance custom_distance = sg.CustomDistance(distance_matrix) # construct embedding based on created distance distance_embedding = converter.embed_distance(custom_distance)