def main(arg):
start = time.time()
source_data = data.Biofile(arg)
sample = source_data.get_header()
feature = source_data.get_index()
X = source_data.get_matrix()
print(X)
print("the frobenis norm of X is {}!".format(np.linalg.norm(X)))
#clustering
component = input("please input the number of cluster you want:")
component = int(component)
print("current avaiable method for nmf(should input in) is \n"
"semi-nmf(snmf),semi-nmf with constraint(csnmf),convex-nmf(cvxnmf) and kernel-nmf(knmf)")
method = input("input the method of nmf you want:")
mx = int(input("input the max iteration numbers:"))
if method not in ["snmf", "csnmf", "cvxnmf", "knmf"]:
raise ValueError("the method you put in is not available now")
if method == "snmf":
print("semi-nmf begins")
F, G, result1 = util.semi_non_negative_factorization(X, n_components = component,max_iter = mx)
if method == "csnmf":
a = input("please input the alpha:")
b = input("please input in beta:")
a, b = float(a), float(b)
F, G, result1 = util.semi_non_negative_factorization_with_straint(X, n_components=component,
alpha = a, beta = b, max_iter=mx)
if method == "cvxnmf":
print("convex-nmf begins")
F, G, result = util.convex_non_negative_factorization(X, n_components= component, max_iter=mx)
if method == "knmf":
k = input("the kernel is:")
p = input("parameter for kernel is:")
p = float(p)
F, G, result = util.kernel_non_negative_factorization(X, n_components = component, max_iter=mx, kernel=k, parameter=p)
#F, G = preprocessing.normalize(F, norm='l2'),np.around(preprocessing.normalize(G, norm='l2'),3)
end = time.time()
G_df = pd.DataFrame(G, index= sample)
G_df.to_csv('G_{}.csv'.format(method))
print (end - start)
print("nmf finished, and the file of probility matrix is generated as csv in current folder")
def test_draw(arg):
start = time.time()
source_data = data.Biofile(arg)
sample = source_data.get_header()
feature = source_data.get_index()
sample_size, feature_size = 60, 9000
#xshape = (106 12042)
X = source_data.get_matrix().T[:sample_size, :feature_size]
group = 16
F, G, result1 = util.semi_non_negative_factorization(X,
n_components=group,
max_iter=100)
print("\n")
F1, G1, result2 = util.semi_non_negative_factorization_with_straint(
X, n_components=group, alpha=0.5, beta=0.5, max_iter=100)
print("\n")
F2, G2, result3 = util.convex_non_negative_factorization(
X, n_components=group, max_iter=100)
print("\n")
F3, G3, result4 = util.kernel_non_negative_factorization(
X, n_components=group, max_iter=100)
def plot_clustering_cvx(arg):
source_data = data.Biofile(arg)
sample = source_data.get_header()
feature = source_data.get_index()
sample_size, feature_size = 106, 12042
#xshape = (106 12042)
X = source_data.get_matrix().T[:sample_size, :feature_size]
cvx_r = util.convex_non_negative_factorization(X.T,
max_iter=100,
n_components=4)
G = cvx_r[1]
sample = sample[:sample_size]
plt.subplot(111)
plt.plot(G[:, 0], G[:, 1], 'co')
plt.title(u'the distribution of items(convex-nmf)')
#items = zip(sample, G)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt.show()
def test_result():
X = np.array([[1.3, 1.8, 4.8, 7.1, 5.0, 5.2, 8.0],
[1.5, 6.9, 3.9, -5.5, -8.5, -3.9, -5.5],
[6.5, 1.6, 8.2, -7.2, -8.7, -7.9, -5.2],
[3.8, 8.3, 4.7, 6.4, 7.5, 3.2, 7.4],
[-7.3, -1.8, -2.1, 2.7, 6.8, 4.8, 6.2]])
print("the Norm of X is {}".format(np.linalg.norm(X)))
F_expected = np.array([[0.05, 0.27], [0.4, -0.4], [0.7, -0.72],
[0.3, 0.08], [-0.51, 0.49]])
Fcvx_expected = np.array([[0.31, 0.53], [0.42, -0.30], [0.56, -0.57],
[0.49, 0.41], [-0.41, 0.36]])
Gt_expected = np.array([[0.61, 0.89, 0.54, 0.77, 0.14, 0.36, 0.84],
[0.12, 0.53, 0.11, 1.03, 0.60, 0.77, 1.16]])
Gcvx_expected = np.array([[0.31, 0.31, 0.29, 0.02, 0, 0, 0.02],
[0, 0.06, 0, 0.31, 0.27, 0.30, 0.36]])
F, G, losses = util.semi_non_negative_factorization(
X, n_components=2, initialization='Kmeans')
F1, G1, losses = util.convex_non_negative_factorization(X, n_components=2)
Gt = G.T
Gt1 = G1.T
#rescale the result
print("\n" + "------------------" * 5 + "\n")
F, Gt = preprocessing.normalize(F, norm='l2'), preprocessing.normalize(
Gt, norm='l2')
F1, Gt1 = preprocessing.normalize(np.dot(
X, F1), norm='l2'), preprocessing.normalize(Gt1, norm='l2')
print("Fsemi_expected:\n {} \n Fcvx_expected : \n {} \n but got \n Fsemi :\n {} \n Fcvnx :\n {}" \
.format(F_expected, Fcvx_expected, np.round(F,2), np.round(F1,2)))
print("-------------------" * 5)
print("Gsemi_expected: \n {} \n Gcvx_expected : \n {} \n but got \n Gsemi :\n {} \n Gcvnx : \n{}" \
.format(Gt_expected,Gcvx_expected, np.round(Gt,2), np.round(Gt1,2)))
def test_plot_kernel(arg):
#it is the method for test of different kernel-nmf
X = np.array([[1.3, 1.8, 4.8, 7.1, 5.0, 5.2, 8.0],
[1.5, 6.9, 3.9, -5.5, -8.5, -3.9, -5.5],
[6.5, 1.6, 8.2, -7.2, -8.7, -7.9, -5.2],
[3.8, 8.3, 4.7, 6.4, 7.5, 3.2, 7.4],
[-7.3, -1.8, -2.1, 2.7, 6.8, 4.8, 6.2]])
"""
source_data = data.Biofile(arg)
sample = source_data.get_header()
feature = source_data.get_index()
sample_size, feature_size = 106, 12042
sample = sample[:sample_size]
#xshape = (106 12042)
print(sample, feature)
X = source_data.get_matrix().T[:sample_size, :feature_size]
"""
semi_r = util.semi_non_negative_factorization(X.T,
max_iter=100,
n_components=2)
semi_r_con = util.semi_non_negative_factorization_with_straint(
X.T, max_iter=100, n_components=2, alpha=0.5, beta=0.5)
semi_r_con1 = util.semi_non_negative_factorization_with_straint(
X.T, max_iter=100, n_components=2, alpha=1, beta=0.5)
semi_r_con2 = util.semi_non_negative_factorization_with_straint(
X.T, max_iter=100, n_components=2, alpha=0.5, beta=1)
cvx_r = util.convex_non_negative_factorization(X.T,
max_iter=60,
n_components=2)
G, G1, G2, G4, G5 = semi_r[1], semi_r_con[1], cvx_r[1], semi_r_con1[
1], semi_r_con2[1]
result, result1, result2, result3, result4 = semi_r[2], semi_r_con[
2], cvx_r[2], semi_r_con1[2], semi_r_con1[2]
x = [i for i in range(30)]
# plot the losses function
plt.title("losses function of {}".format(arg[:-4]))
plt.xlabel("iteration times")
plt.ylabel("losses")
plt.plot(x, result[:30], 'r', marker='.', label='sNMF')
plt.plot(x, result1[:30], 'b', marker='.', label='con-sNMF(0.5, 0.5)')
plt.plot(x, result3[:30], 'c', marker='.', label='con-sNMF(0, 0.5)')
plt.plot(x, result4[:30], 'm', marker='.', label='con-sNMF(0.5, 1)')
plt.plot(x, result2[:30], 'y', marker='.', label='cvx-NMF')
plt.legend(bbox_to_anchor=[1, 1])
plt.grid()
plt.show()
plt.savefig("figure1.png", dpi=300)
plt.close()
#plot the clustering result
plt1 = plt
plt1.subplot(221)
plt1.plot(G[:, 0], G[:, 1], 'ro')
plt1.title(u'the distribution of items(sNMF)')
#items = zip(sample, G)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.subplot(222)
plt1.plot(G1[:, 0], G1[:, 1], 'bo')
plt1.title(u'the distribution of items(sNMF(0.5, 0.5))')
#items = zip(sample, G1)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.subplot(223)
plt1.plot(G4[:, 0], G4[:, 1], 'co')
plt1.title(u'the distribution of items(sNMF(0, 0.5))')
#items = zip(sample, G4)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.subplot(224)
plt1.plot(G2[:, 0], G2[:, 1], 'mo')
plt1.title(u'the distribution of items(convexNMF)')
#items = zip(sample, G2)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.show()
def test_plot(arg):
"""
X = np.array([[1.3, 1.8, 4.8, 7.1, 5.0, 5.2, 8.0],
[1.5, 6.9, 3.9, -5.5, -8.5, -3.9, -5.5],
[6.5, 1.6, 8.2, -7.2, -8.7, -7.9, -5.2],
[3.8, 8.3, 4.7, 6.4, 7.5, 3.2, 7.4],
[-7.3, -1.8, -2.1, 2.7, 6.8, 4.8, 6.2]
])
"""
source_data = data.Biofile(arg)
sample = source_data.get_header()
feature = source_data.get_index()
sample_size, feature_size = 106, 12042
sample = sample[:sample_size]
#xshape = (106 12042)
print(sample, feature)
X = source_data.get_matrix().T[:sample_size, :feature_size]
mx = 100
labs = ['rbf', 'poly', 'sigmoid']
semi_r = util.kernel_non_negative_factorization(X.T,
n_components=2,
max_iter=mx,
parameter=100) #rbf 0.5
semi_r_con = util.kernel_non_negative_factorization(X.T,
n_components=2,
max_iter=mx,
kernel='poly',
parameter=0.5) #ploy 2
semi_r_con1 = util.kernel_non_negative_factorization(
X.T, n_components=2, max_iter=mx, kernel='sigmoid',
parameter=0.1) #sigmoid 0.5
semi_r_con2 = util.convex_non_negative_factorization(X.T,
max_iter=mx,
n_components=2)
#semi_r = util.semi_non_negative_factorization_with_straint(X.T, max_iter = mx,n_components=2 ,initialization= 'Kmeans',alpha = 0.01, beta = 0.01)
#semi_r_con = util.semi_non_negative_factorization_with_straint(X.T, max_iter=mx,n_components=2 ,initialization= 'Kmeans',alpha= 10, beta = 10)
#semi_r_con1 = util.semi_non_negative_factorization_with_straint(X.T, max_iter=mx,n_components=2, initialization= 'Kmeans',alpha= 0, beta = 10)
#semi_r_con2 = util.semi_non_negative_factorization_with_straint(X.T, max_iter=mx,n_components=2, initialization= 'Kmeans',alpha= 10, beta = 0)
#convex_r_con = util.convex_non_negative_factorization(X.T, n_components=2, max_iter=mx)
G, G1, G2, G3 = semi_r[1], semi_r_con[1], semi_r_con1[1], semi_r_con2[1]
result, result1, result2, result3 = semi_r[2], semi_r_con[2], semi_r_con1[
2], semi_r_con2[2]
x = [i for i in range(mx)]
# plot the losses function
plt.title("losses function of {}".format(arg[:-4]))
plt.xlabel("iteration times")
plt.ylabel("losses")
plt.plot(x, result[:mx], 'r', marker='.', label='kNMF({})'.format(labs[0]))
plt.plot(x,
result1[:mx],
'b',
marker='.',
label='kNMF({})'.format(labs[1]))
plt.plot(x,
result2[:mx],
'c',
marker='.',
label='kNMF({})'.format(labs[2]))
plt.plot(x, result3[:mx], 'm', marker='.', label='cvxnmf')
"""
plt.plot(x, result[:mx], 'r', marker = '.', label = 'sNMF')
plt.plot(x, result1[:mx], 'b', marker ='.' , label = 'sNMF(0.5,0.5)')
plt.plot(x, result2[:mx], 'c', marker ='.', label = 'sNMF(0,0.5)')
plt.plot(x, result3[:mx], 'm', marker ='.', label = 'sNMF(0.5,1)')
plt.plot(x, result4[:mx], 'k', marker = '.', label = 'cvx-NMF')
"""
plt.legend(bbox_to_anchor=[1, 1])
plt.grid()
plt.show()
#plot the clustering result
plt1 = plt
plt1.subplot(221)
plt1.plot(G[:, 0], G[:, 1], 'ro')
plt1.title(u'the distribution of items(knmf({}))'.format(labs[0]))
#items = zip(sample, G)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.subplot(222)
plt1.plot(G1[:, 0], G1[:, 1], 'bo')
plt1.title(u'the distribution of items(knmf({}))'.format(labs[1]))
#items = zip(sample, G1)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.subplot(223)
plt1.plot(G2[:, 0], G2[:, 1], 'co')
plt1.title(u'the distribution of items((knmf({}))'.format(labs[2]))
#items = zip(sample, G4)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.subplot(224)
plt1.plot(G3[:, 0], G3[:, 1], 'mo')
plt1.title(u'the distribution of items(convex-nmf))')
#items = zip(sample, G2)
#for item in items:
# item_name, item_data = item[0], item[1]
# plt1.text(item_data[0], item_data[1], item_name,
# horizontalalignment='center',
# verticalalignment='top')
plt1.show()