def do_nmf(self,verbose=False):
## Filter out the low frequency ones.
fidx = []
fwords = []
for i,w in enumerate(self.idx2word):
if self.lexical_counts[w] >= self.min_count_nmf:
fidx.append(i)
fwords.append(w)
fidx.append(self.alphabet_size)
fwords.append(self.start_symbol)
self.nmf = nmf.NMF(self.unigram_features[fidx,:], fwords, ssf=0)
## This may not be the start symbol of the grammar.
start_idx = len(fidx)-1
self.nmf.start(start_idx)
## What does this even mean then?
self.kernels = [self.start_symbol]
assert not self.start_symbol in self.word2idx
self.nmf.excluded.add(start_idx)
while len(self.kernels) < self.nonterminals:
a, ai = self.nmf.find_but_dont_add()
if verbose:
print("Adding kernel", a, "count", self.lexical_counts[a])
self.nmf.add_basis(ai)
self.kernels.append(a)
self.nmf.initialise_frank_wolfe()
def test_nmf():
print('Testing NMF on synthetic data...')
alpha = 1
beta = 10
n_components = 2
shape = (8, 6)
n = nmf.NMF(shape=shape, n_components=n_components, alpha=alpha, beta=beta)
n_est = nmf.NMF(shape=n.shape,
n_components=n.n_components,
alpha=alpha,
beta=beta)
elbo = n_est.fit(n.X, max_iter=2000)
for i in range(len(elbo)):
if i == 0:
print(elbo[0])
else:
print('{} ; {}'.format(elbo[i], elbo[i] - elbo[i - 1]))
print('Num. iterations: {}'.format(len(elbo) - 1))
print('done.')
def main():
path = r'G:\School\thesis\DS4_T\wav_renamed'
settings = "settings.set"
WavFiles = glob.glob(path + '\\*.' + 'wav')
for index in WavFiles:
spec = readwav(index)
nmfObject = nmf.NMF(spec[2], settings)
nmfObject.printparas()
nmfObject.decompose()
nmfObject.writeresults("test1")
os.system("pause")
#show(spec)
return
def main():
V = np.array([[1, 1, 1, 1, 1], [0, 1, 0, 1, 0], [0, 1, 0, 1, 0]])
r = 2
W, H = nmf.NMF(V, r, lamb=0, maxit=2000)
Ve = np.dot(W, H)
np.set_printoptions(precision=2, suppress=True)
for i in range(r):
Hi = np.zeros((r, np.shape(H)[1]))
Hi[i, :] = H[i, :]
TMP = np.dot(W, Hi)
pl.imsave("V" + str(i) + ".png", TMP, cmap=pl.cm.gray)
pl.imsave("V.png", V, cmap=pl.cm.gray)
pl.imsave("W.png", W, cmap=pl.cm.gray)
pl.imsave("H.png", H, cmap=pl.cm.gray)
pl.imsave("Ve.png", Ve, cmap=pl.cm.gray)
return 0
def init_nmf_with_kernels(self, kernels): ## Given known clusters initialise the NMF so we can do the Frank-Wolfe estimation ## for the unary rules. fidx = [] fwords = [] for i,w in enumerate(self.idx2word): fidx.append(i) fwords.append(w) fidx.append(self.alphabet_size) fwords.append(self.start_symbol) self.nmf = nmf.NMF(self.unigram_features[fidx,:], fwords, ssf=0) ## This may not be the start symbol of the grammar. start_idx = len(fidx)-1 self.nmf.start(start_idx) self.kernels = [ kernels[0] ] for a in kernels[1:]: ai = self.word2idx[a] self.nmf.add_basis(ai) self.kernels.append(a) self.nmf.initialise_frank_wolfe()
with open('mnist.txt', 'r') as mFr:
for line in mFr:
a = line.split(',')
b = []
for item in a:
b.append(float(item))
categoryM.append(b[-1])
mDataSet.append(b)
mFr.close()
calculate.calculate(kMeans.kMeans(gDataSet, 2), categoryG, 2)
calculate.calculate(kMeans.kMeans(mDataSet, 10), categoryM, 10)
calculate.calculate(nmf.NMF(gDataSet, 2), categoryG, 2)
calculate.calculate(nmf.NMF(mDataSet, 10), categoryG, 10)
calculate.calculate(spectral.spectral(gDataSet, 2, 3), categoryG, 2)
calculate.calculate(spectral.spectral(gDataSet, 2, 6), categoryG, 2)
calculate.calculate(spectral.spectral(gDataSet, 2, 9), categoryG, 2)
calculate.calculate(spectral.spectral(mDataSet, 10, 3), categoryM, 10)
calculate.calculate(spectral.spectral(mDataSet, 10, 6), categoryM, 10)
calculate.calculate(spectral.spectral(mDataSet, 10, 9), categoryM, 10)
def _nmf():
return nmf.NMF(files='texts')
#W_v = np.eye(6700)
n = len(X)
m = len(X[0])
#X = np.array(X)
W_u = np.array(W_u)
W_v = np.array(W_v)
# Initialize the W_u & W_v
D_u = np.zeros(shape=(n, n))
D_v = np.zeros(shape=(m, m))
sum_W_u = np.sum(W_u, axis=1)
sum_W_v = np.sum(W_v, axis=1)
for i in range(n):
D_u[i][i] = sum_W_u[i]
for h in range(m):
D_v[h][h] = sum_W_v[h]
U = np.random.rand(n, k)
H = np.random.rand(k, k)
V = np.random.rand(m, k)
U_final, H_final, V_final = nmf.NMF(X, U, H, V, D_u, W_u, D_v, W_v)
print("\nU_final:\n\n", U_final)
print("\nH_final:\n\n", H_final)
print("\nV_final:\n\n", V_final)
df = pd.read_pickle('articles.pkl')
content = df.content.as_matrix()
content = df.content.as_matrix()
n_topics = 7
# vectorize our content
vector = TfidfVectorizer(max_features=5000, stop_words='english')
vector_matrix = vector.fit_transform(content).toarray()
features = np.array(vector.get_feature_names())
our_nmf = nmf.NMF(vector_matrix, n_topics, max_iter=100)
W, H = our_nmf.fit()
### Using Your NMF Function
print "MSR: {}".format(our_nmf.msr())
for row in H:
ind = row.argsort()[:-10 - 1:-1]
print features[ind]
### Built-In NMF
print "\nsklearn.NMF\n"
skl_nmf = sklearn.decomposition.NMF(n_components=n_topics,
random_state=1,