def plot_gtm(config, pca_data, labels, ids):
regul = 0.1
niter = 1000
gtm = ugtm.runGTM(
data=pca_data,
k=config.k,
m=config.m,
s=config.rbf_width_factor,
regul=regul,
niter=niter,
doPCA=False,
n_components=config.pca_n_components,
missing=config.missing,
missing_strategy=config.missing_strategy,
random_state=config.random_state,
verbose=config.verbose,
)
gtm.plot_html(
labels=labels,
ids=ids,
discrete=config.discrete_labels,
output=config.output + "_control_plot",
cname=config.color_map,
pointsize=config.pointsize,
alpha=config.alpha,
title="Control Graph (All Samples)",
prior=config.gtm_prior,
do_interpolate=config.interpolate,
)
return None
def test_runGTM(self):
gtm = ugtm.runGTM(data=self.data, k=self.k)
self.assertEqual(gtm.converged, True)
self.assertEqual(gtm.matR.shape, (self.n_individuals, self.n_nodes))
self.assertAlmostEqual(sum(gtm.matR[0, :]), 1.0)
self.assertEqual(gtm.matMeans.shape, (self.n_individuals, 2))
self.assertEqual(gtm.matModes.shape, (self.n_individuals, 2))
def test_transform(self):
gtm = ugtm.runGTM(data=self.data, k=self.k, doPCA=True)
transformed = ugtm.transform(
optimizedModel=gtm, train=self.data, test=self.data, doPCA=True)
self.assertEqual(transformed.converged, True)
np.testing.assert_almost_equal(gtm.matR, transformed.matR, decimal=7)
np.testing.assert_almost_equal(
gtm.matMeans, transformed.matMeans, decimal=7)
def setUp(self):
np.random.seed(0)
self.n_dimensions = 10
self.n_individuals = 100
self.n_nodes = 4
self.k = 2
self.n_rbf_centers = 4
self.m = 2
self.data = np.random.randn(self.n_individuals, self.n_dimensions)
self.labels = np.random.choice([1, 2], size=self.n_individuals)
self.gtm = ugtm.runGTM(self.data, k=self.k, m=self.m)
def test_runWorkflow(self):
train = np.random.randn(20, 10)
test = np.random.randn(20, 10)
labels = np.random.choice(["class1", "class2"], size=20)
activity = np.random.randn(20)
gtm = ugtm.runGTM(train)
gtm.write("tests/output_tests/testout1")
gtm_coordinates = gtm.matMeans
gtm_modes = gtm.matModes
gtm_responsibilities = gtm.matR
gtm.plot_multipanel(output="tests/output_tests/testout2",
labels=labels,
discrete=True,
pointsize=20)
gtm.plot_multipanel(output="tests/output_tests/testout3",
labels=activity,
discrete=False,
pointsize=20)
gtm.plot_multipanel(output="tests/output_tests/testout4",
labels=labels,
discrete=True,
pointsize=20,
do_interpolate=False)
gtm.plot_multipanel(output="tests/output_tests/testout5",
labels=activity,
discrete=False,
pointsize=20,
do_interpolate=False)
gtm.plot(output="tests/output_tests/testout6", pointsize=20)
gtm.plot(output="tests/output_tests/testout7",
labels=labels,
discrete=True,
pointsize=20)
gtm.plot(output="tests/output_tests/testout8",
labels=activity,
discrete=False,
pointsize=20)
gtm.plot_html(output="tests/output_tests/testout9", pointsize=20)
gtm.plot_html(output="tests/output_tests/testout10",
labels=activity,
discrete=False,
pointsize=20)
gtm.plot_html(output="tests/output_tests/testout11",
labels=labels,
discrete=True,
pointsize=20)
gtm.plot_html(output="tests/output_tests/testout12",
labels=activity,
discrete=False,
pointsize=20,
do_interpolate=False,
ids=labels)
gtm.plot_html(output="tests/output_tests/testout13",
labels=labels,
discrete=True,
pointsize=20,
do_interpolate=False)
transformed = ugtm.transform(optimizedModel=gtm,
train=train,
test=test)
transformed.plot_html(output="tests/output_tests/testout14",
pointsize=20)
transformed.plot(output="tests/output_tests/testout15", pointsize=20)
gtm.plot_html_projection(output="tests/output_tests/testout16",
projections=transformed,
labels=labels,
discrete=True,
pointsize=20)
predicted_labels = ugtm.GTC(train=train, test=test, labels=labels)
predicted_model = ugtm.advancedGTC(train=train,
test=test,
labels=labels)
predicted_model = ugtm.advancedGTC(train=train,
test=test,
labels=labels,
doPCA=True,
n_components=-1)
ugtm.printClassPredictions(predicted_model,
"tests/output_tests/testout17")
predicted = ugtm.GTR(train=train, test=test, labels=activity)
predicted = ugtm.GTC(train=train, test=test, labels=labels)
ugtm.crossvalidateGTC(data=train,
labels=labels,
s=1,
regul=1,
n_repetitions=10,
n_folds=5)
ugtm.crossvalidateGTR(data=train, labels=activity, s=1, regul=1)
ugtm.crossvalidatePCAC(data=train, labels=labels, n_neighbors=7)
ugtm.crossvalidateSVCrbf(data=train, labels=labels, C=1, gamma=1)
ugtm.crossvalidateSVCrbf(data=train, labels=labels, C=1)
ugtm.crossvalidateSVR(data=train, labels=activity, C=1, epsilon=1)
ugtm.crossvalidatePCAR(data=train, labels=activity, n_neighbors=7)
gtm = ugtm.runkGTM(train, doKernel=True, kernel="linear")
output=args.output,
cname=args.cname,
pointsize=args.pointsize,
alpha=args.alpha)
np.savetxt(args.output + "_tsne.csv", data_r, delimiter=',')
exit
# if it's for GTM visualization
elif args.model == 'GTM':
start = time.time()
gtm = ugtm.runGTM(data=data,
k=k,
m=m,
s=s,
l=l,
niter=niter,
doPCA=args.pca,
n_components=args.n_components,
missing=args.missing,
missing_strategy=args.missing_strategy,
random_state=args.random_state,
verbose=args.verbose)
print("k:%s, m:%s, l:%s, s:%s" % (k, m, l, s))
end = time.time()
elapsed = end - start
print("time taken for GTM: ", elapsed)
np.savetxt(args.output + "_matmeans.csv", gtm.matMeans, delimiter=',')
gtm.plot_multipanel(labels=labels,
output=args.output,
discrete=discrete,
cname=args.cname,
pointsize=args.pointsize,