def test_als(self):
""" Test accuracy of alternating least-squares NMF algorithm
against the MATLAB-computed version
"""
# set data and initializing constants
keys = [array([i + 1]) for i in range(4)]
data_local = array([[1.0, 2.0, 6.0], [1.0, 3.0, 0.0], [1.0, 4.0, 6.0],
[5.0, 1.0, 4.0]])
data = self.sc.parallelize(zip(keys, data_local))
h0 = array([[0.09082617, 0.85490047, 0.57234593],
[0.82766740, 0.21301186, 0.90913979]])
# if the rows of h are not normalized on each iteration:
h_true = array([[0., 0.6010, 0.9163], [0.8970, 0.1556, 0.7423]])
w_true = array([[4.5885, 1.5348], [1.3651, 0.2184], [5.9349, 1.0030],
[0., 5.5147]])
# if the columns of h are normalized (as in the current implementation):
scale_mat = diag(norm(h_true, axis=1))
h_true = dot(LinAlg.inv(scale_mat), h_true)
w_true = dot(w_true, scale_mat)
# calculate NMF using the Thunder implementation
# (maxiter=9 corresponds with Matlab algorithm)
nmf_thunder = NMF(k=2, method="als", h0=h0, maxiter=9)
nmf_thunder.calc(data)
h_thunder = nmf_thunder.h
w_thunder = array(nmf_thunder.w.values().collect())
tol = 1e-03 # allow small error
assert (allclose(w_thunder, w_true, atol=tol))
assert (allclose(h_thunder, h_true, atol=tol))
def test_als(self):
""" Test accuracy of alternating least-squares NMF algorithm
against the MATLAB-computed version
"""
# set data and initializing constants
keys = [array([i + 1]) for i in range(4)]
data_local = array([[1.0, 2.0, 6.0], [1.0, 3.0, 0.0], [1.0, 4.0, 6.0],
[5.0, 1.0, 4.0]])
data = self.sc.parallelize(zip(keys, data_local))
h0 = array([[0.09082617, 0.85490047, 0.57234593],
[0.82766740, 0.21301186, 0.90913979]])
w0_local = array([[0.40967962, 0.82131499], [0.98889171, 0.54361205],
[0.08074117, 0.44954983], [0.68648657, 0.71598381]])
w0 = self.sc.parallelize(zip(keys, w0_local))
# if the rows of h are not normalized on each iteration:
h_true = array([[0.88909291, 0., 0.45868362],
[1.54548542, 6.33549752, 10.65274926]])
w_true = array([[0.67478533, 0.47374843], [0.08587805, 0.12829046],
[0., 0.58218579], [5.34371512, 0.14884191]])
# if the columns of h are normalized (as in the current implementation):
scale_mat = diag(norm(h_true, axis=1))
h_true = dot(LinAlg.inv(scale_mat), h_true)
w_true = dot(w_true, scale_mat)
# calculate NMF using the Thunder implementation
nmf_thunder = NMF(k=2, method="als", h0=h0, w0=w0, maxiter=20)
nmf_thunder.calc(data)
h_thunder = nmf_thunder.h
w_thunder = array(nmf_thunder.w.values().collect())
tol = 1e-04 # allow small error
assert (allclose(w_thunder, w_true, atol=tol))
assert (allclose(h_thunder, h_true, atol=tol))
def test_init(self):
"""
test performance of whole function, including random initialization
"""
data_local = array([[1.0, 2.0, 6.0], [1.0, 3.0, 0.0], [1.0, 4.0, 6.0],
[5.0, 1.0, 4.0]])
data = self.sc.parallelize(
zip([array([i]) for i in range(data_local.shape[0])], data_local))
nmf_thunder = NMF(k=2, recon_hist='final')
nmf_thunder.calc(data)
# check to see if Thunder's solution achieves close-to-optimal reconstruction error
# scikit-learn's solution achieves 2.993952
# matlab's non-deterministic implementation usually achieves < 2.9950 (when it converges)
assert (nmf_thunder.recon_err < 2.9950)
def test_als(self):
""" Test accuracy of alternating least-squares NMF algorithm
against the MATLAB-computed version
"""
# set data and initializing constants
keys = [array([i+1]) for i in range(4)]
data_local = array([
[1.0, 2.0, 6.0],
[1.0, 3.0, 0.0],
[1.0, 4.0, 6.0],
[5.0, 1.0, 4.0]])
data = self.sc.parallelize(zip(keys, data_local))
h0 = array(
[[0.09082617, 0.85490047, 0.57234593],
[0.82766740, 0.21301186, 0.90913979]])
w0_local = array(
[[0.40967962, 0.82131499],
[0.98889171, 0.54361205],
[0.08074117, 0.44954983],
[0.68648657, 0.71598381]])
w0 = self.sc.parallelize(zip(keys, w0_local))
# if the rows of h are not normalized on each iteration:
h_true = array(
[[0.88909291, 0. , 0.45868362],
[1.54548542, 6.33549752, 10.65274926]])
w_true = array(
[[0.67478533, 0.47374843],
[0.08587805, 0.12829046],
[0. , 0.58218579],
[5.34371512, 0.14884191]])
# if the columns of h are normalized (as in the current implementation):
scale_mat = diag(norm(h_true, axis=1))
h_true = dot(LinAlg.inv(scale_mat), h_true)
w_true = dot(w_true, scale_mat)
# calculate NMF using the Thunder implementation
nmf_thunder = NMF(k=2, method="als", h0=h0, w0=w0, maxiter=20)
nmf_thunder.calc(data)
h_thunder = nmf_thunder.h
w_thunder = array(nmf_thunder.w.values().collect())
tol = 1e-04 # allow small error
assert(allclose(w_thunder, w_true, atol=tol))
assert(allclose(h_thunder, h_true, atol=tol))
def test_init(self):
"""
test performance of whole function, including random initialization
"""
data_local = array([
[1.0, 2.0, 6.0],
[1.0, 3.0, 0.0],
[1.0, 4.0, 6.0],
[5.0, 1.0, 4.0]])
data = self.sc.parallelize(zip([array([i]) for i in range(data_local.shape[0])], data_local))
nmf_thunder = NMF(k=2)
nmf_thunder.calc(data)
# check to see if Thunder's solution achieves close-to-optimal reconstruction error
# scikit-learn's solution achieves 2.993952
# matlab's non-deterministic implementation usually achieves < 2.9950 (when it converges)
assert(nmf_thunder.rec_err < 2.9950)
def test_als(self):
""" Test accuracy of alternating least-squares NMF algorithm
against the MATLAB-computed version
"""
# set data and initializing constants
keys = [array([i+1]) for i in range(4)]
data_local = array([
[1.0, 2.0, 6.0],
[1.0, 3.0, 0.0],
[1.0, 4.0, 6.0],
[5.0, 1.0, 4.0]])
data = self.sc.parallelize(zip(keys, data_local))
h0 = array(
[[0.09082617, 0.85490047, 0.57234593],
[0.82766740, 0.21301186, 0.90913979]])
# if the rows of h are not normalized on each iteration:
h_true = array(
[[0. , 0.6010, 0.9163],
[0.8970, 0.1556, 0.7423]])
w_true = array(
[[4.5885, 1.5348],
[1.3651, 0.2184],
[5.9349, 1.0030],
[0. , 5.5147]])
# if the columns of h are normalized (as in the current implementation):
scale_mat = diag(norm(h_true, axis=1))
h_true = dot(LinAlg.inv(scale_mat), h_true)
w_true = dot(w_true, scale_mat)
# calculate NMF using the Thunder implementation
# (maxiter=9 corresponds with Matlab algorithm)
nmf_thunder = NMF(k=2, method="als", h0=h0, maxiter=9)
nmf_thunder.calc(data)
h_thunder = nmf_thunder.h
w_thunder = array(nmf_thunder.w.values().collect())
tol = 1e-03 # allow small error
assert(allclose(w_thunder, w_true, atol=tol))
assert(allclose(h_thunder, h_true, atol=tol))
