def main():
# load data
mat = scipy.io.loadmat('../data/COIL20.mat')
X = mat['X'] # data
X = X.astype(float)
y = mat['Y'] # label
y = y[:, 0]
# construct affinity matrix
kwargs = {"metric": "euclidean", "neighborMode": "knn", "weightMode": "heatKernel", "k": 5, 't': 1}
W = construct_W.construct_W(X, **kwargs)
num_fea = 100 # specify the number of selected features
num_cluster = 20 # specify the number of clusters, it is usually set as the number of classes in the ground truth
# obtain the feature weight matrix
Weight = MCFS.mcfs(X, n_selected_features=num_fea, W=W, n_clusters=20)
# sort the feature scores in an ascending order according to the feature scores
idx = MCFS.feature_ranking(Weight)
# obtain the dataset on the selected features
selected_features = X[:, idx[0:num_fea]]
# perform kmeans clustering based on the selected features and repeats 20 times
nmi_total = 0
acc_total = 0
for i in range(0, 20):
nmi, acc = unsupervised_evaluation.evaluation(X_selected=selected_features, n_clusters=num_cluster, y=y)
nmi_total += nmi
acc_total += acc
# output the average NMI and average ACC
print 'NMI:', float(nmi_total)/20
print 'ACC:', float(acc_total)/20
def featureSelection(listNewTimeSeries, numFeature, numCluster=5):
"""
@description : select features using MCFS algorithem.
---------
@param : numFeature -- how many features are to be selected.
numCluster -- parameter required in MFCS, deafault is set to 5.
-------
@Returns : selected_features -- selected features
idx -- the indexes of selected features in original feature set.
-------
"""
kwargs = {
"metric": "euclidean",
"neighborMode": "knn",
"weightMode": "heatKernel",
"k": 5,
't': 1
}
listNewTimeSeries = np.array(listNewTimeSeries)
W = construct_W.construct_W(listNewTimeSeries, **kwargs)
Weight = MCFS.mcfs(listNewTimeSeries,
n_selected_features=numFeature,
W=W,
n_clusters=numCluster)
idx = MCFS.feature_ranking(Weight)
selected_features = listNewTimeSeries[:, idx[0:numFeature]]
return selected_features, idx
def mcfs(trnin, num_fea):
from skfeature.utility import construct_W
kwargs_W = {
"metric": "euclidean",
"neighbor_mode": "knn",
"weight_mode": "heat_kernel",
"k": 5,
't': 1
}
W = construct_W.construct_W(trnin, **kwargs_W)
from skfeature.function.sparse_learning_based import MCFS
score = MCFS.mcfs(trnin, num_fea, W=W)
idx = MCFS.feature_ranking(score)
selfea = idx[0:num_fea]
return selfea
def main():
# load data
mat = scipy.io.loadmat('../data/COIL20.mat')
X = mat['X'] # data
X = X.astype(float)
y = mat['Y'] # label
y = y[:, 0]
# construct affinity matrix
kwargs = {
"metric": "euclidean",
"neighborMode": "knn",
"weightMode": "heatKernel",
"k": 5,
't': 1
}
W = construct_w.construct_w(X, **kwargs)
num_fea = 100 # specify the number of selected features
num_cluster = 20 # specify the number of clusters, it is usually set as the number of classes in the ground truth
# obtain the feature weight matrix
Weight = MCFS.mcfs(X, n_selected_features=num_fea, W=W, n_clusters=20)
# sort the feature scores in an ascending order according to the feature scores
idx = MCFS.feature_ranking(Weight)
# obtain the dataset on the selected features
selected_features = X[:, idx[0:num_fea]]
# perform kmeans clustering based on the selected features and repeats 20 times
nmi_total = 0
acc_total = 0
for i in range(0, 20):
nmi, acc = unsupervised_evaluation.evaluation(
X_selected=selected_features, n_clusters=num_cluster, y=y)
nmi_total += nmi
acc_total += acc
# output the average NMI and average ACC
print('NMI:', old_div(float(nmi_total), 20))
print('ACC:', old_div(float(acc_total), 20))
def SKF_mcfs(X, y):
# construct affinity matrix
kwargs = {
"metric": "euclidean",
"neighborMode": "knn",
"weightMode": "heatKernel",
"k": 5,
't': 1
}
W = construct_W(X, **kwargs)
num_fea = X.shape[1] # specify the number of selected features
num_cluster = len(
set(y)
) # specify the number of clusters, it is usually set as the number of classes in the ground truth
# obtain the feature weight matrix
Weight = MCFS.mcfs(X,
n_selected_features=num_fea,
W=W,
n_clusters=num_cluster)
return MCFS.feature_ranking(Weight)
def MCFS_FS(X_train, k):
# construct affinity matrix
kwargs = {
"metric": "euclidean",
"neighborMode": "knn",
"weightMode": "heatKernel",
"k": 5,
't': 1
}
W = construct_W.construct_W(X_train, **kwargs)
num_fea_ = k # specify the number of selected features
num_cluster = 20 # specify the number of clusters, it is usually set as the number of classes in the ground truth
# obtain the feature weight matrix
Weight = MCFS.mcfs(X_train,
n_selected_features=num_fea_,
W=W,
n_clusters=20)
# sort the feature scores in an ascending order according to the feature scores
idx = MCFS.feature_ranking(Weight)
return (idx, Weight)
def mcfs_score(diheds):
import scipy.io
import numpy
from numpy import mean
import os
#os.chdir('/home/anu/Downloads/scikit-feature-1.0.0')
from skfeature.function.sparse_learning_based import MCFS
from skfeature.utility import construct_W
from skfeature.utility import unsupervised_evaluation
idx = []
kwargs = {"metric": "euclidean", "neighborMode": "knn", "weightMode": "heatKernel", "k": 5, 't': 1}
#change the path for every system to be run.
#os.chdir('/home/anu/Downloads/DESRES-Trajectory_GTT-1-protein/GTT-1-protein')
for i in range(0,len(diheds),5):
X= diheds[i]
W = construct_W.construct_W(X, **kwargs)
score = MCFS.mcfs(X, n_selected_features=20, W=W, n_clusters=20)
idx.append(score)
col_mean = mean(idx, axis =0)
imp_features=MCFS.feature_ranking(col_mean)
return col_mean,imp_features
def calc_MCFS(data, n_features, n_clusters=20):
kwargs_W = {
"metric": "euclidean",
"neighbor_mode": "knn",
"weight_mode": "heat_kernel",
"k": 5,
't': 1
}
W = construct_W.construct_W(data, **kwargs_W)
return MCFS.mcfs(data,
n_selected_features=n_features,
W=W,
n_clusters=n_clusters).max(1)
def MCFS(X, y=None, **kwargs):
# construct affinity matrix
kwargs = {
"metric": "euclidean",
"neighborMode": "knn",
"weightMode": "heatKernel",
"k": 5,
't': 1
}
W = construct_W.construct_W(X, **kwargs)
num_cluster = len(np.unique(y))
# obtain the feature weight matrix
Weight = MCFS_CLASS.mcfs(X,
n_selected_features=X.shape[1],
W=W,
n_clusters=num_cluster)
# sort the feature scores in an ascending order according to the feature scores
idx = MCFS_CLASS.feature_ranking(Weight)
return idx
def select(dataset, features_number, clusters_number):
app_logger.info(
'STARTED [MCFS Selection] on {0} with features number = {1}'.format(
dataset, features_number),
extra=LOGGER_EXTRA_OBJECT)
# Retrieving all feature extracted by tsfresh from the pickles on the disk
current_dir = os.getcwd().split('\\')[-1]
projet_dir = 'MCFS-Unsupervisioned-Feature-Selection'
if current_dir == projet_dir:
all_features_train = pd.read_pickle(
'Pickle/AllFeatures/Train/{0}.pkl'.format(dataset))
all_features_test = pd.read_pickle(
'Pickle/AllFeatures/Test/{0}.pkl'.format(dataset))
else:
all_features_train = pd.read_pickle(
'../Pickle/AllFeatures/Train/{0}.pkl'.format(dataset))
all_features_test = pd.read_pickle(
'../Pickle/AllFeatures/Test/{0}.pkl'.format(dataset))
app_logger.info(
'All features (including target column) trainset shape: {0}'.format(
all_features_train.shape),
extra=LOGGER_EXTRA_OBJECT)
app_logger.info(
'All features (including target column) testset shape: {0}'.format(
all_features_test.shape),
extra=LOGGER_EXTRA_OBJECT)
# np.savetxt(r'testDataFrame.txt', all_features_test.values, fmt='%d')
# Retrieving indipendent columns of both set and known labels of the test set
indipendent_columns_train = all_features_train.iloc[:, 1:]
indipendent_columns_test = all_features_test.iloc[:, 1:]
known_labels_test = all_features_test.iloc[:, 0]
# Building matrix W for MCFS algorithm
kwargs = {
'metric': 'euclidean',
'neighbor_mode': 'knn',
'weight_mode': 'binary',
'k': 3
# 'weight_mode': 'heat_kernel',
# 'k': 5,
# 't': 1
}
W = construct_W.construct_W(indipendent_columns_train.values, **kwargs)
# MCFS gives a weight to each features
kwargs = {'W': W, 'n_clusters': clusters_number}
weighted_features = MCFS.mcfs(indipendent_columns_train.values,
features_number, **kwargs)
# Ordering the features according to their weight
ordered_features = MCFS.feature_ranking(weighted_features)
# Getting only the first 'features_number' features
selected_features = ordered_features[0:features_number]
# Getting names of selected features
names_selected_features = []
for feature_index in selected_features:
names_selected_features.append(
indipendent_columns_train.columns[feature_index])
# Selected only the selected features on the train set
selected_features_train = indipendent_columns_train.loc[:,
names_selected_features]
app_logger.info('Selected features trainset: {0}'.format(
selected_features_train.shape),
extra=LOGGER_EXTRA_OBJECT)
# Selected only the selected features on the test set
selected_features_test = indipendent_columns_test.loc[:,
names_selected_features]
app_logger.info('Selected features testset: {0}'.format(
selected_features_test.shape),
extra=LOGGER_EXTRA_OBJECT)
'''
# Pickles for rfd
if selected_features_train.shape[0] > 1000:
print('Test-set')
selected_features_test.to_pickle('../rfd/Pickle_rfd/MCFS/{0}.pkl'.format(dataset))
else:
print('Train-set')
selected_features_train.to_pickle('../rfd/Pickle_rfd/MCFS/{0}.pkl'.format(dataset))
exit()
'''
# Running k-means according to selected features
test_feature_selection.testFeatureSelectionWithRepeatedKMeans(
'MCFS', features_number, dataset, selected_features_train.values,
selected_features_test.values, clusters_number, known_labels_test)
app_logger.info('ENDED [MCFS Selection] on {0}'.format(dataset),
extra=LOGGER_EXTRA_OBJECT)
# Testing
#select('TwoPatterns', 10, 4)
def generate_result_dist(dataset, x,y,num_select, zero_mean=False, N=1000, t=0.6, thresh=0.1):
if zero_mean == False:
x = normalize(x,axis=0)
else:
x = standardize_feature(x)
n,d = x.shape
if num_select==300:
start_dim = 20; step = 20
elif num_select==200: # the dimension
start_dim = 20; step = 10
elif num_select==100:
start_dim = 10; step = 10
elif num_select==50:
start_dim = 10; step = 5
elif num_select == 20:
start_dim = 4; step = 2
else:
start_dim = 5; step = 1
dimension_list = list(range(start_dim,num_select+1,step))
######### rank: parameter preserve_pctg, num_use #########
D0 = compute_dist(x)
preserve_pctg_list = [0.2,0.4,0.6,0.8,1] #dimension 0
num_use_list = [0.1,0.2,0.3,0.4,0.5] #dimension 1
rank_result = np.zeros([len(preserve_pctg_list),len(num_use_list),7,len(dimension_list)])
rank_result_l1 = np.zeros([len(preserve_pctg_list),len(num_use_list),7,len(dimension_list)])
rank_result_l2 = np.zeros([len(preserve_pctg_list),len(num_use_list),7,len(dimension_list)])
rank_result_lmax = np.zeros([len(preserve_pctg_list),len(num_use_list),7,len(dimension_list)])
for i,preserve_pctg in enumerate(preserve_pctg_list):
for j,num_use in enumerate(num_use_list):
print(i,j)
rank_selected, rank_selected_l1, rank_selected_l2, rank_selected_lmax= ranking_selection(x, num_select, N=N, num_use=int(num_use*d+1),sample_pctg=1, preserve_pctg=preserve_pctg)
rank_selected = list(rank_selected)[::-1]
for k,dimension in enumerate(dimension_list): #performance using different number fo features
s = rank_selected[:dimension]
rank_x = x[:,s]
D_rank = compute_dist(rank_x)
rank_result[i,j,0,k] = ef.dif_dist(D0,D_rank,'l1')
rank_result[i,j,1,k] = ef.dif_dist(D0,D_rank,'l2')
rank_result[i,j,2,k] = ef.dif_dist(D0,D_rank,'lmax')
s_l1 = rank_selected_l1[:dimension]
rank_l1_x = x[:,s_l1]
D1 = compute_dist(rank_l1_x)
rank_result_l1[i,j,0,k] = ef.dif_dist(D0,D1,'l1')
rank_result_l1[i,j,1,k] = ef.dif_dist(D0,D1,'l2')
rank_result_l1[i,j,2,k] = ef.dif_dist(D0,D1,'lmax')
s_l2 = rank_selected_l2[:dimension]
rank_l2_x = x[:,s_l2]
D2 = compute_dist(rank_l2_x)
rank_result_l2[i,j,0,k] = ef.dif_dist(D0,D2,'l1')
rank_result_l2[i,j,1,k] = ef.dif_dist(D0,D2,'l2')
rank_result_l2[i,j,2,k] = ef.dif_dist(D0,D2,'lmax')
s_lmax = rank_selected_lmax[:dimension]
rank_lmax_x = x[:,s_lmax]
D_max = compute_dist(rank_lmax_x)
rank_result_lmax[i,j,0,k] = ef.dif_dist(D0,D_max,'l1')
rank_result_lmax[i,j,1,k] = ef.dif_dist(D0,D_max,'l2')
rank_result_lmax[i,j,2,k] = ef.dif_dist(D0,D_max,'lmax')
np.save('./result/'+dataset+'/rank_dist',rank_result)
np.save('./result/'+dataset+'/rank_l1_dist',rank_result_l1)
np.save('./result/'+dataset+'/rank_l2_dist',rank_result_l2)
np.save('./result/'+dataset+'/rank_lmax_dist',rank_result_lmax)
######## lap_score ###########
lap_score_result = np.zeros([7,len(dimension_list)])
lap_score_selected = lap_score.lap_score(x)
lap_score_selected = list(np.argsort(lap_score_selected)[:num_select]) #find minimum
for k,dimension in enumerate(dimension_list): #performance using different number fo features
s = lap_score_selected[:dimension]
lap_score_x = x[:,s]
D1 = compute_dist(lap_score_x)
lap_score_result[0,k] = ef.dif_dist(D0,D1,'l1')
lap_score_result[1,k] = ef.dif_dist(D0,D1,'l2')
lap_score_result[2,k] = ef.dif_dist(D0,D1,'lmax')
np.save('./result/'+dataset+'/lap_score_dist',lap_score_result)
######## SPEC ###########
SPEC_result = np.zeros([7,len(dimension_list)])
SPEC_selected = SPEC.spec(x)
SPEC_selected = list(np.argsort(SPEC_selected)[:num_select]) #find minimum
for k,dimension in enumerate(dimension_list): #performance using different number fo features
s = SPEC_selected[:dimension]
SPEC_x = x[:,s]
D1 = compute_dist(SPEC_x)
SPEC_result[0,k] = ef.dif_dist(D0,D1,'l1')
SPEC_result[1,k] = ef.dif_dist(D0,D1,'l2')
SPEC_result[2,k] = ef.dif_dist(D0,D1,'lmax')
np.save('./result/'+dataset+'/SPEC_dist',SPEC_result)
####### MCFS parameter: num_clusters ##############
num_clusters_list = [5,10,20,30]
MCFS_result = np.zeros([len(num_clusters_list),7,len(dimension_list)])
for i,num_clusters in enumerate(num_clusters_list):
MCFS_W = MCFS.mcfs(x,num_select,**{'n_clusters':num_clusters})
MCFS_selected = [np.max(np.abs(x)) for x in MCFS_W] #find maximum
MCFS_selected= np.argsort(MCFS_selected)[-num_select:]
MCFS_selected = list(MCFS_selected)[::-1]
for k,dimension in enumerate(dimension_list): #performance using different number fo features
s = MCFS_selected[:dimension]
MCFS_x = x[:,s]
D1 = compute_dist(MCFS_x)
MCFS_result[i,0,k] = ef.dif_dist(D0,D1,'l1')
MCFS_result[i,1,k] = ef.dif_dist(D0,D1,'l2')
MCFS_result[i,2,k] = ef.dif_dist(D0,D1,'lmax')
np.save('./result/'+dataset+'/MCFS_dist',MCFS_result)
return rank_result, rank_result_l1, rank_result_l2,rank_result_lmax,lap_score_result, SPEC_result, MCFS_result
def compare_methods(x,y,num_select,pctg=0.5,sample_pctg=1, num_clusters=5,zero_mean=False,dim=1,t=0.8,thresh=0.1):
if zero_mean == False:
x = normalize(x,axis=0)
else:
x = standardize_feature(x)
n,d = x.shape
# idx = np.random.permutation(n)
# x,y = x[idx], y[idx]
#
# ######### split train and test #########
# X=x;Y=y
# train_num = int(n*0.6)
# test_num = n-int(n*0.6)
# x=X[:train_num,:]; y=Y[:train_num]
# x_test = X[-test_num:,:];y_test = Y[-test_num:]
########### calculate ######################
start_time = time.clock()
rf_result = random_selection(x, num_select, N=500, num_use=int(0.5*d),pctg=pctg, two_sided=False)
print('rf running time:',time.clock()-start_time)
start_time = time.clock()
rank_result,l1,l2,lmax= ranking_selection(x, num_select, N=500, num_use=int(0.5*d),sample_pctg=1, preserve_pctg=pctg)
print('rank running time:',time.clock()-start_time)
start_time = time.clock()
lap_score_result = lap_score.lap_score(x)
lap_score_result= np.argsort(lap_score_result)[:num_select] #find minimum
print('lap_score running time:',time.clock()-start_time)
start_time = time.clock()
SPEC_result = SPEC.spec(x)
print('SPEC running time:',time.clock()-start_time)
SPEC_result= np.argsort(SPEC_result)[:num_select] #find minimum
'''sparse learning based'''
start_time = time.clock()
MCFS_W = MCFS.mcfs(x,num_select,**{'n_clusters':num_clusters})
print('MCFS running time:',time.clock()-start_time)
MCFS_result = [np.max(np.abs(x)) for x in MCFS_W] #find maximum
MCFS_result= np.argsort(MCFS_result)[-num_select:]
# start_time = time.clock()
# NDFS_W = NDFS.ndfs(x,**{'n_clusters':num_clusters})
# print('NDFS running time:',time.clock()-start_time)
# NDFS_result = [np.sqrt(np.sum(x**2)) for x in NDFS_W] #find maximum
# NDFS_result= np.argsort(NDFS_result)[-num_select:]
#
# start_time = time.clock()
# UDFS_W = UDFS.udfs(x,**{'n_clusters':num_clusters})
# print('UDFS running time:',time.clock()-start_time)
# UDFS_result = [np.sqrt(np.sum(x**2)) for x in UDFS_W] #find minimum ??????????????????????
# UDFS_result= np.argsort(UDFS_result)[:num_select]
# prop_x = x[:,list(stepwise)]
rf_x = x[:,list(rf_result)]
rank_x = x[:,list(rank_result)]
l1_x = x[:,list(l1)]
l2_x = x[:,list(l2)]
lmax_x = x[:,list(lmax)]
lap_score_x = x[:,list(lap_score_result)]
SPEC_x = x[:,list(SPEC_result)]
MCFS_x = x[:,list(MCFS_result)]
# NDFS_x = x[:,list(NDFS_result)]
# UDFS_x = x[:,list(UDFS_result)]
# '''[KNN purity NMI dgm0 dgm1], each one is a matrix'''
# methods = ['rf','rank','lap_score','SPEC','MCFS']
# for method in methods:
# if method=='rf':
# selected_feature = list(rf_result).reverse()
# elif method=='rank':
# selected_feature = list(rank_result).reverse()
# elif method=='lap_score':
# selected_feature = list(lap_score_result)
# elif method=='SPEC':
# selected_feature = list(SPEC_result)
# else:
# selected_feature = list(MCFS_result).reverse()
#
# if num_select<=50: # the dimension
# start_dim = 5; step = 2
# else:
# start_dim = 10; step = 5
print('KNN accuracy')
print('rf', ef.knn_accuracy(x,y,rf_result))
print('rank', ef.knn_accuracy(x,y,rank_result))
print('l1', ef.knn_accuracy(x,y,l1))
print('l2', ef.knn_accuracy(x,y,l2))
print('lmax', ef.knn_accuracy(x,y,lmax))
print('lap_score', ef.knn_accuracy(x,y,lap_score_result))
print('SPEC', ef.knn_accuracy(x,y,SPEC_result))
print('MCFS',ef.knn_accuracy(x,y,MCFS_result))
# print('NDFS',ef.knn_accuracy(x_test,y_test,NDFS_result))
# print('UDFS',ef.knn_accuracy(x_test,y_test,UDFS_result),'\n')
# print('connectivity')
# print('rf', ef.connectivity(x,rf_x,pctg, two_sided))
# print('rank', ef.connectivity(x,rank_x,pctg, two_sided))
# print('lap_score', ef.connectivity(x,lap_score_x,pctg, two_sided))
# print('SPEC', ef.connectivity(x,SPEC_x,pctg, two_sided))
# print('cut-SPEC', ef.connectivity(x,CSPEC_x,pctg, two_sided))
# print('MCFS',ef.connectivity(x,MCFS_x,pctg, two_sided))
# print('NDFS',ef.connectivity(x,NDFS_x,pctg, two_sided))
# print('UDFS',ef.connectivity(x,UDFS_x,pctg, two_sided),'\n')
print('purity score | NMI')
print('origin', ef.purity_score(x,y))
print('rf', ef.purity_score(rf_x,y))
print('rank', ef.purity_score(rank_x,y))
print('lap_score', ef.purity_score(lap_score_x,y))
print('SPEC', ef.purity_score(SPEC_x,y) )
print('MCFS', ef.purity_score(MCFS_x,y))
dgm = ef.compute_dgm(x, t, dim, thresh)
dgm_rf = ef.compute_dgm(rf_x, t, dim, thresh)
dgm_rank = ef.compute_dgm(rank_x, t, dim, thresh)
dgm_l1 = ef.compute_dgm(l1_x, t, dim, thresh)
dgm_l2 = ef.compute_dgm(l2_x, t, dim, thresh)
dgm_lmax = ef.compute_dgm(lmax_x, t, dim, thresh)
dgm_lap_score = ef.compute_dgm(lap_score_x, t, dim, thresh)
dgm_SPEC = ef.compute_dgm(SPEC_x, t, dim, thresh)
dgm_MCFS = ef.compute_dgm(MCFS_x, t, dim, thresh)
# plt.figure()
# plt.plot(dgm[:,-2:], 'ro')
# plt.figure()
# plt.plot(dgm_rf[:,-2:], 'ro')
# plt.figure()
# plt.plot(dgm_rank[:,-2:], 'ro')
# plt.figure()
# plt.plot(dgm_SPEC[:,-2:], 'ro')
# plt.figure()
# plt.plot(dgm_MCFS[:,-2:], 'ro')
print('dgm distance')
print('rf', ef.dgm_distance(dgm,dgm_rf,'W', dim),' ',ef.dgm_distance(dgm,dgm_rf,'B', dim))
print('rank', ef.dgm_distance(dgm,dgm_rank,'W', dim),' ',ef.dgm_distance(dgm,dgm_rank,'B', dim))
print('l1', ef.dgm_distance(dgm,dgm_l1,'W', dim),' ',ef.dgm_distance(dgm,dgm_l1,'B', dim))
print('l2', ef.dgm_distance(dgm,dgm_l2,'W', dim),' ',ef.dgm_distance(dgm,dgm_l2,'B', dim))
print('lmax', ef.dgm_distance(dgm,dgm_lmax,'W', dim),' ',ef.dgm_distance(dgm,dgm_lmax,'B', dim))
print('lap_score', ef.dgm_distance(dgm,dgm_lap_score,'W', dim),' ',ef.dgm_distance(dgm,dgm_lap_score,'B', dim))
print('SPEC', ef.dgm_distance(dgm,dgm_SPEC,'W', dim),' ',ef.dgm_distance(dgm,dgm_SPEC,'B', dim))
print('MCFS', ef.dgm_distance(dgm,dgm_MCFS,'W', dim),' ',ef.dgm_distance(dgm,dgm_MCFS,'B', dim))
def compare_methods(x,
y,
num_select,
pctg=0.1,
pack_size=1,
num_clusters=5,
two_sided=False):
n, d = x.shape
idx = np.random.permutation(n)
x, y = x[idx], y[idx]
######### split train and test #########
X = x
Y = y
train_num = int(n * 0.7)
test_num = n - int(n * 0.7)
x = X[:train_num, :]
y = Y[:train_num]
x_test = X[-test_num:, :]
y_test = Y[-test_num:]
########### other methods ######################
''' Similarity based: lap_score SPEC '''
start_time = time.clock()
lap_score_result = lap_score.lap_score(x)
lap_score_result = np.argsort(lap_score_result)[:num_select]
print('lap_score running time:', time.clock() - start_time)
# _,stepwise = backward_distance_selection(x,num_select,pctg,pack_size) #pctg controls sensitivity to outliers
start_time = time.clock()
rf_result = random_selection(x,
num_select,
N=300,
num_use=int(d / 2),
pctg=pctg,
two_sided=two_sided)
print('rf running time:', time.clock() - start_time)
start_time = time.clock()
SPEC_result = SPEC.spec(x)
print('SPEC running time:', time.clock() - start_time)
SPEC_result = np.argsort(SPEC_result)[:num_select] #find minimum
start_time = time.clock()
CSPEC_result = cut_spec(x, pctg=0.15)
print('cut-SPEC running time:', time.clock() - start_time)
CSPEC_result = np.argsort(CSPEC_result)[:num_select] #find minimum
'''sparse learning based'''
start_time = time.clock()
MCFS_W = MCFS.mcfs(x, num_select)
print('MCFS running time:', time.clock() - start_time)
MCFS_result = [np.max(np.abs(x)) for x in MCFS_W] #find maximum
MCFS_result = np.argsort(MCFS_result)[-num_select:]
# start_time = time.clock()
# NDFS_W = NDFS.ndfs(x,**{'n_clusters':num_clusters})
# print('NDFS running time:',time.clock()-start_time)
# NDFS_result = [np.sqrt(np.sum(x**2)) for x in NDFS_W] #find maximum
# NDFS_result= np.argsort(NDFS_result)[-num_select:]
#
# start_time = time.clock()
# UDFS_W = UDFS.udfs(x,**{'n_clusters':num_clusters})
# print('UDFS running time:',time.clock()-start_time)
# UDFS_result = [np.sqrt(np.sum(x**2)) for x in UDFS_W] #find minimum ??????????????????????
# UDFS_result= np.argsort(UDFS_result)[:num_select]
# prop_x = x[:,list(stepwise)]
rf_x = x[:, list(rf_result)]
lap_score_x = x[:, list(lap_score_result)]
SPEC_x = x[:, list(SPEC_result)]
CSPEC_x = x[:, list(CSPEC_result)]
MCFS_x = x[:, list(MCFS_result)]
# NDFS_x = x[:,list(NDFS_result)]
# UDFS_x = x[:,list(UDFS_result)]
print('\n')
print('Class Seperability')
# print('prop', ef.class_seperability(prop_x,y))
print('rf', ef.class_seperability(rf_x, y))
print('lap_score', ef.class_seperability(lap_score_x, y))
print('SPEC', ef.class_seperability(SPEC_x, y))
print('cut-SPEC', ef.class_seperability(CSPEC_x, y))
print('MCFS', ef.class_seperability(MCFS_x, y))
# print('NDFS',ef.class_seperability(NDFS_x,y))
# print('UDFS',ef.class_seperability(UDFS_x,y))
print('\n')
print('KNN accuracy')
# print('prop', ef.knn_accuracy(prop_x,y))
print('rf', ef.knn_accuracy(x_test, y_test, rf_result))
print('lap_score', ef.knn_accuracy(x_test, y_test, lap_score_result))
print('SPEC', ef.knn_accuracy(x_test, y_test, SPEC_result))
print('cut-SPEC', ef.knn_accuracy(x_test, y_test, CSPEC_result))
print('MCFS', ef.knn_accuracy(x_test, y_test, MCFS_result))
# print('NDFS',ef.knn_accuracy(x_test,y_test,NDFS_result))
# print('UDFS',ef.knn_accuracy(x_test,y_test,UDFS_result),'\n')
print('\n')
print('connectivity')
# print('prop', ef.knn_accuracy(prop_x,y))
print('rf', ef.connectivity(x, rf_x, pctg, two_sided))
print('lap_score', ef.connectivity(x, lap_score_x, pctg, two_sided))
print('SPEC', ef.connectivity(x, SPEC_x, pctg, two_sided))
print('cut-SPEC', ef.connectivity(x, CSPEC_x, pctg, two_sided))
print('MCFS', ef.connectivity(x, MCFS_x, pctg, two_sided))
data = np.loadtxt("./data/GaussianTopologyNode.txt")
edge = np.loadtxt("./data/GaussianTopologyEdge.txt")
timeStart = datetime.datetime.now()
if useEdge:
W = ConstructWbyEdge.ConstructWbyEdge(data, edge, t=1)
else:
kwrags_W = {
"metric": "euclidean",
"neighbor_mode": "knn",
"weight_mode": "heat_kernel",
"k": 3,
"t": 1
}
W = construct_W(data, **kwrags_W)
result = MCFS.mcfs(data, n_selected_features=2, W=W, n_clusters=2)
print result
timeEnd = datetime.datetime.now()
print "Run Time: ", timeEnd - timeStart
elif dataSet == 1:
data = np.loadtxt("./data/SwissRollTopologyNode.txt")
edge = np.loadtxt("./data/SwissRollTopologyEdge.txt")
timeStart = datetime.datetime.now()
if useEdge:
W = ConstructWbyEdge.ConstructWbyEdge(data, edge, t=1)
else:
kwrags_W = {
"metric": "euclidean",
def mcfs(train_set, test_set, features_number, clusters_number):
# Features to delete
features_to_delete = []
for i in range(3, len(sys.argv)):
features_to_delete.append(sys.argv[i])
# Retrieving indipendent columns of both set and known labels of the test set
indipendent_columns_train = train_set.iloc[:, 1:]
indipendent_columns_test = test_set.iloc[:, 1:]
known_labels_test = test_set.iloc[:, 0]
# Building matrix W for MCFS algorithm
kwargs = {
'metric': 'euclidean',
'neighbor_mode': 'knn',
'weight_mode': 'binary',
'k': 3
}
W = construct_W.construct_W(indipendent_columns_train.values, **kwargs)
# MCFS gives a weight to each features
kwargs = {'W': W, 'n_clusters': clusters_number}
weighted_features = MCFS.mcfs(indipendent_columns_train.values,
features_number, **kwargs)
# Ordering the features according to their weight
ordered_features = MCFS.feature_ranking(weighted_features)
# Getting only the first 'features_number' features
selected_features = ordered_features[0:features_number]
# Getting names of selected features
names_selected_features = []
for feature_index in selected_features:
names_selected_features.append(
indipendent_columns_train.columns[feature_index])
# Deleting "feature to delete"
names_selected_features = [
feature for feature in names_selected_features
if feature not in features_to_delete
]
if len(names_selected_features
) != len(selected_features) - len(features_to_delete):
kmeans_rfd_logger.error(
'One or more feature "to delete" is/are not correct.')
else:
# Selected only the selected features on the train set
selected_features_train = indipendent_columns_train.loc[:,
names_selected_features]
# Selected only the selected features on the test set
selected_features_test = indipendent_columns_test.loc[:,
names_selected_features]
kmeans_rfd_logger.info(
'(Deleted features: {0})'.format(features_to_delete))
# Running k-means according to selected features
run_kmeans(len(names_selected_features),
selected_features_train.values,
selected_features_test.values, clusters_number,
known_labels_test)
