def main():
if len(argv) != 6:
print_usage(basename(argv[0]))
exit(1)
print_log = True
n_components = int(argv[1])
input_train_data = load(argv[2])
input_test_data = load(argv[3])
output_train_data_file = argv[4]
output_test_data_file = argv[5]
t = time()
pca = ProbabilisticPCA(n_components=n_components)
output_train_data = pca.fit_transform(input_train_data)
t = time()-t
if print_log:
print 'PCA fitting to {0} with {1} components to save (with transformation): {2[0]:d}m {2[1]:02d}s'.format(input_train_data.shape, n_components, (int(t//60), int(t%60)) )
print 'New shape: {}'.format(output_train_data.shape)
t = time()
output_test_data = pca.transform (input_test_data)
t = time()-t
if print_log:
print 'PCA transformation of {0} with {1} components to save: {2[0]:d}m {2[1]:02d}s'.format(input_test_data.shape, n_components, (int(t//60), int(t%60)) )
print 'New shape: {}'.format(output_test_data.shape)
save(output_train_data_file, output_train_data)
save(output_test_data_file, output_test_data)
def test_probabilistic_pca_vs_pca():
"""Test that PCA matches ProbabilisticPCA with homoscedastic=True
"""
n, p = 100, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
pca = PCA(n_components=2).fit(X)
ppca = ProbabilisticPCA(n_components=2).fit(X)
assert_array_almost_equal(pca.score_samples(X), ppca.score(X))
def test_probabilistic_pca_vs_pca():
"""Test that PCA matches ProbabilisticPCA with homoscedastic=True
"""
n, p = 100, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
pca = PCA(n_components=2).fit(X)
ppca = ProbabilisticPCA(n_components=2).fit(X)
assert_array_almost_equal(pca.score_samples(X), ppca.score(X))
def test_probabilistic_pca_2():
"""Test that probabilistic PCA correctly separated different datasets"""
n, p = 100, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
ppca = ProbabilisticPCA(n_components=2)
ppca.fit(X)
ll1 = ppca.score(X)
ll2 = ppca.score(rng.randn(n, p) * .2 + np.array([3, 4, 5]))
assert_greater(ll1.mean(), ll2.mean())
def test_probabilistic_pca_1():
"""Test that probabilistic PCA yields a reasonable score"""
n, p = 1000, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
ppca = ProbabilisticPCA(n_components=2)
ppca.fit(X)
ll1 = ppca.score(X)
h = -0.5 * np.log(2 * np.pi * np.exp(1) * 0.1 ** 2) * p
np.testing.assert_almost_equal(ll1.mean() / h, 1, 0)
def test_probabilistic_pca_1():
"""Test that probabilistic PCA yields a reasonable score"""
n, p = 1000, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
ppca = ProbabilisticPCA(n_components=2)
ppca.fit(X)
ll1 = ppca.score(X)
h = -0.5 * np.log(2 * np.pi * np.exp(1) * 0.1**2) * p
np.testing.assert_almost_equal(ll1.mean() / h, 1, 0)
def test_probabilistic_pca_4():
"""Check that ppca select the right model"""
n, p = 200, 3
rng = np.random.RandomState(0)
Xl = rng.randn(n, p) + rng.randn(n, 1) * np.array([3, 4, 5]) + np.array([1, 0, 7])
Xt = rng.randn(n, p) + rng.randn(n, 1) * np.array([3, 4, 5]) + np.array([1, 0, 7])
ll = np.zeros(p)
for k in range(p):
ppca = ProbabilisticPCA(n_components=k)
ppca.fit(Xl)
ll[k] = ppca.score(Xt).mean()
assert_true(ll.argmax() == 1)
def test_probabilistic_pca_4():
"""Check that ppca select the right model"""
n, p = 200, 3
rng = np.random.RandomState(0)
Xl = (rng.randn(n, p) + rng.randn(n, 1) * np.array([3, 4, 5]) +
np.array([1, 0, 7]))
Xt = (rng.randn(n, p) + rng.randn(n, 1) * np.array([3, 4, 5]) +
np.array([1, 0, 7]))
ll = np.zeros(p)
for k in range(p):
ppca = ProbabilisticPCA(n_components=k)
ppca.fit(Xl)
ll[k] = ppca.score(Xt).mean()
assert_true(ll.argmax() == 1)
def pcaImgWrdMat(highDim, lowDim):
(options, args) = parser.parse_args(sys.argv[1:]) #@UnusedVariable
dataset = options.dataset
# method = options.method
#acquire the category list
catmap = getCatMap(dataset)
catList = catmap.keys()
# the number of categories in category list
# nCategory = len(catList)
for catName in catList:
print '%s : %d : %d\n' % (catName, highDim, lowDim)
catPosFileName = rootDir + dataset + iwmDir + catName + str(
highDim) + iwmext
catPosData = np.loadtxt(catPosFileName, dtype=np.int, delimiter=' ')
nPosImages = catPosData.shape[0]
catNegFileName = rootDir + dataset + iwmDir + 'NEG' + catName + str(
highDim) + iwmext
catNegData = np.loadtxt(catNegFileName, dtype=np.int, delimiter=' ')
nNegImages = catNegData.shape[0]
catData = np.vstack((catPosData, catNegData))
labels = np.vstack((np.ones(
(nPosImages, 1), np.int), np.zeros((nNegImages, 1), np.int)))
print 'pca...'
pcaData = PCA(n_components=lowDim).fit(catData).transform(catData)
pcaData = np.hstack((pcaData, labels))
pcaDataFileName = rootDir + dataset + outputDir + catName + str(
highDim) + str(lowDim) + '.pca'
np.savetxt(pcaDataFileName, pcaData, fmt='%f', delimiter=' ')
print 'ppca...'
ppcaData = ProbabilisticPCA(
n_components=lowDim).fit(catData).transform(catData)
ppcaData = np.hstack((ppcaData, labels))
ppcaDataFileName = rootDir + dataset + outputDir + catName + str(
highDim) + str(lowDim) + '.ppca'
np.savetxt(ppcaDataFileName, ppcaData, fmt='%f', delimiter=' ')
print 'rpca...'
rpcaData = RandomizedPCA(
n_components=lowDim).fit(catData).transform(catData)
rpcaData = np.hstack((rpcaData, labels))
rpcaDataFileName = rootDir + dataset + outputDir + catName + str(
highDim) + str(lowDim) + '.rpca'
np.savetxt(rpcaDataFileName, rpcaData, fmt='%f', delimiter=' ')
print 'kpca...'
kpcaData = KernelPCA(
n_components=lowDim).fit(catData).transform(catData)
kpcaData = np.hstack((kpcaData, labels))
kpcaDataFileName = rootDir + dataset + outputDir + catName + str(
highDim) + str(lowDim) + '.kpca'
np.savetxt(kpcaDataFileName, kpcaData, fmt='%f', delimiter=' ')
print 'spca...'
spcaData = MiniBatchSparsePCA(
n_components=lowDim, n_iter=100).fit(catData).transform(catData)
spcaData = np.hstack((spcaData, labels))
spcaDataFileName = rootDir + dataset + outputDir + catName + str(
highDim) + str(lowDim) + '.spca'
np.savetxt(spcaDataFileName, spcaData, fmt='%f', delimiter=' ')
pass
def test_probabilistic_pca_3():
"""The homoscedastic model should work slightly worse
than the heteroscedastic one in over-fitting condition
"""
n, p = 100, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
ppca = ProbabilisticPCA(n_components=2)
ppca.fit(X)
ll1 = ppca.score(X)
ppca.fit(X, homoscedastic=False)
ll2 = ppca.score(X)
def test_probabilistic_pca_3():
"""The homoscedastic model should work slightly worse
than the heteroscedastic one in over-fitting condition
"""
n, p = 100, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
ppca = ProbabilisticPCA(n_components=2)
ppca.fit(X)
ll1 = ppca.score(X)
ppca.fit(X, homoscedastic=False)
ll2 = ppca.score(X)
def test_probabilistic_pca_2():
"""Test that probabilistic PCA correctly separated different datasets"""
n, p = 100, 3
rng = np.random.RandomState(0)
X = rng.randn(n, p) * .1 + np.array([3, 4, 5])
ppca = ProbabilisticPCA(n_components=2)
ppca.fit(X)
ll1 = ppca.score(X)
ll2 = ppca.score(rng.randn(n, p) * .2 + np.array([3, 4, 5]))
assert_greater(ll1.mean(), ll2.mean())
