From fb7b686b144cc62322f5615ed403c4b7c22bc27e Mon Sep 17 00:00:00 2001
From: Yannick DAYER <yannick.dayer@idiap.ch>
Date: Tue, 7 Dec 2021 17:44:19 +0100
Subject: [PATCH] Added GMM dask tests, renamed tests
---
bob/learn/em/test/test_gmm.py | 186 ++++++++++++++++++++++++----------
1 file changed, 130 insertions(+), 56 deletions(-)
diff --git a/bob/learn/em/test/test_gmm.py b/bob/learn/em/test/test_gmm.py
index c11d5e1..785b0ae 100644
--- a/bob/learn/em/test/test_gmm.py
+++ b/bob/learn/em/test/test_gmm.py
@@ -101,7 +101,7 @@ def test_GMMStats():
# Clean-up
os.unlink(filename)
-def test_GMMMachine_1():
+def test_GMMMachine():
# Test a GMMMachine basic features
weights = np.array([0.5, 0.5], "float64")
@@ -182,9 +182,41 @@ def test_GMMMachine_1():
assert (gmm == gmm6) is False
assert gmm.is_similar_to(gmm6) is False
+ # Saving and loading
+ filename = tempfile.mkstemp(suffix=".hdf5")[1]
+ gmm.save(HDF5File(filename, "w"))
+ gmm1 = GMMMachine.from_hdf5(HDF5File(filename, "r"))
+ assert gmm == gmm1
+ gmm.save(filename)
+ gmm1 = GMMMachine.from_hdf5(filename)
+ assert gmm == gmm1
+ os.unlink(filename)
+
+ # Weights
+ n_gaussians = 5
+ machine = GMMMachine(n_gaussians)
+
+ default_weights = np.full(shape=(n_gaussians,), fill_value=1.0 / n_gaussians)
+ default_log_weights = np.full(
+ shape=(n_gaussians,), fill_value=np.log(1.0 / n_gaussians)
+ )
+
+ # Test weights getting and setting
+ np.testing.assert_almost_equal(machine.weights, default_weights)
+ np.testing.assert_almost_equal(machine.log_weights, default_log_weights)
-def test_GMMMachine_2():
- # Test a GMMMachine (statistics)
+ modified_weights = default_weights
+ modified_weights[: n_gaussians // 2] = (1 / n_gaussians) / 2
+ modified_weights[n_gaussians // 2 + n_gaussians % 2 :] = (1 / n_gaussians) * 1.5
+
+ # Ensure setter works (log_weights is updated correctly)
+ machine.weights = modified_weights
+ np.testing.assert_almost_equal(machine.weights, modified_weights)
+ np.testing.assert_almost_equal(machine.log_weights, np.log(modified_weights))
+
+
+def test_GMMMachine_stats():
+ """Tests a GMMMachine (statistics)"""
arrayset = load_array(resource_filename("bob.learn.em", "data/faithful.torch3_f64.hdf5"))
gmm = GMMMachine(n_gaussians=2)
@@ -206,8 +238,8 @@ def test_GMMMachine_2():
np.testing.assert_almost_equal(stats.sum_pxx, stats_ref.sum_pxx, decimal=10)
-def test_GMMMachine_3():
- # Test a GMMMachine (log-likelihood computation)
+def test_GMMMachine_ll_computation():
+ """Test a GMMMachine (log-likelihood computation)"""
data = load_array(resource_filename("bob.learn.em", "data/data.hdf5"))
gmm = GMMMachine(n_gaussians=2)
@@ -215,13 +247,12 @@ def test_GMMMachine_3():
gmm.means = load_array(resource_filename("bob.learn.em", "data/means.hdf5"))
gmm.variances = load_array(resource_filename("bob.learn.em", "data/variances.hdf5"))
- # Compare the log-likelihood with the one obtained using Chris Matlab
- # implementation
+ # Compare the log-likelihood with the one obtained using Chris Matlab implementation
matlab_ll_ref = -2.361583051672024e+02
np.testing.assert_almost_equal(gmm.log_likelihood(data), matlab_ll_ref, decimal=10)
-def test_GMMMachine_4():
+def test_GMMMachine_single_ll_vs_multiple():
np.random.seed(3) # FIXING A SEED
@@ -241,7 +272,7 @@ def test_GMMMachine_4():
assert np.isclose(ll, gmm.log_likelihood(data).mean())
-def test_GMMStats_2():
+def test_GMMStats_operations():
"""Test a GMMStats."""
# Initializing a GMMStats
data = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9], [7, 8, 9]])
@@ -319,18 +350,17 @@ def test_machine_parameters():
np.testing.assert_equal(machine.variances, np.ones((n_gaussians, n_features)))
# Setters
-
new_means = np.repeat([[1], [2], [3]], n_features, axis=1)
machine.means = new_means
assert machine.means.shape == (n_gaussians, n_features)
- np.testing.assert_almost_equal(machine.means, new_means)
+ np.testing.assert_equal(machine.means, new_means)
new_variances = np.repeat([[0.2], [1.1], [1]], n_features, axis=1)
machine.variances = new_variances
assert machine.variances.shape == (n_gaussians, n_features)
- np.testing.assert_almost_equal(machine.variances, new_variances)
+ np.testing.assert_equal(machine.variances, new_variances)
-def test_kmeans_plusplus_init():
+def test_gmm_kmeans_plusplus_init():
n_gaussians = 3
machine = GMMMachine(
n_gaussians,
@@ -344,7 +374,7 @@ def test_kmeans_plusplus_init():
np.testing.assert_almost_equal(machine.variances, expected_variances)
-def test_kmeans_parallel_init():
+def test_gmm_kmeans_parallel_init():
n_gaussians = 3
machine = GMMMachine(
n_gaussians,
@@ -567,9 +597,6 @@ def loadGMM():
return gmm
-def equals(x, y, epsilon):
- return (abs(x - y) < epsilon).all()
-
def test_gmm_ML_1():
"""Trains a GMMMachine with ML_GMMTrainer"""
ar = load_array(resource_filename("bob.learn.em", "data/faithful.torch3_f64.hdf5"))
@@ -632,44 +659,6 @@ def test_gmm_ML_2():
np.testing.assert_allclose(gmm.weights, weightsML_ref, atol=1e-4)
-def test_gmm_ML_parallel():
- """Trains a GMMMachine with ML_GMMTrainer; compares to a reference"""
-
- ar = da.array(load_array(resource_filename("bob.learn.em", "data/dataNormalized.hdf5")))
-
- # Initialize GMMMachine
- gmm = GMMMachine(n_gaussians=5)
- gmm.means = load_array(resource_filename("bob.learn.em", "data/meansAfterKMeans.hdf5")).astype("float64")
- gmm.variances = load_array(resource_filename("bob.learn.em", "data/variancesAfterKMeans.hdf5")).astype("float64")
- gmm.weights = np.exp(load_array(resource_filename("bob.learn.em", "data/weightsAfterKMeans.hdf5")).astype("float64"))
-
- threshold = 0.001
- gmm.variance_thresholds = threshold
-
- # Initialize ML Trainer
- gmm.mean_var_update_threshold = 0.001
- gmm.max_fitting_steps = 25
- gmm.convergence_threshold = 0.00001
- gmm.update_means = True
- gmm.update_variances = True
- gmm.update_weights = True
-
- # Run ML
- gmm.fit(ar)
-
- # Test results
- # Load torch3vision reference
- meansML_ref = load_array(resource_filename("bob.learn.em", "data/meansAfterML.hdf5"))
- variancesML_ref = load_array(resource_filename("bob.learn.em", "data/variancesAfterML.hdf5"))
- weightsML_ref = load_array(resource_filename("bob.learn.em", "data/weightsAfterML.hdf5"))
-
- # Compare to current results
- np.testing.assert_allclose(gmm.means, meansML_ref, atol=3e-3)
- np.testing.assert_allclose(gmm.variances, variancesML_ref, atol=3e-3)
- np.testing.assert_allclose(gmm.weights, weightsML_ref, atol=1e-4)
-
-
-
def test_gmm_MAP_1():
"""Train a GMMMachine with MAP_GMMTrainer"""
ar = load_array(resource_filename("bob.learn.em", "data/faithful.torch3_f64.hdf5"))
@@ -788,7 +777,7 @@ def test_gmm_MAP_3():
def test_gmm_test():
- """ Tests a GMMMachine by computing scores against a model and comparing to a reference
+ """Tests a GMMMachine by computing scores against a model and comparing to a reference
"""
ar = load_array(resource_filename("bob.learn.em", "data/dataforMAP.hdf5"))
@@ -810,3 +799,88 @@ def test_gmm_test():
# Compare current results to torch3vision
assert abs(score-score_mean_ref)/score_mean_ref<1e-4
+
+
+def test_gmm_ML_dask():
+ """Trains a GMMMachine with dask array data; compares to a reference"""
+
+ ar = da.array(load_array(resource_filename("bob.learn.em", "data/dataNormalized.hdf5")))
+
+ # Initialize GMMMachine
+ gmm = GMMMachine(n_gaussians=5)
+ gmm.means = load_array(resource_filename("bob.learn.em", "data/meansAfterKMeans.hdf5")).astype("float64")
+ gmm.variances = load_array(resource_filename("bob.learn.em", "data/variancesAfterKMeans.hdf5")).astype("float64")
+ gmm.weights = np.exp(load_array(resource_filename("bob.learn.em", "data/weightsAfterKMeans.hdf5")).astype("float64"))
+
+ threshold = 0.001
+ gmm.variance_thresholds = threshold
+
+ # Initialize ML Trainer
+ gmm.mean_var_update_threshold = 0.001
+ gmm.max_fitting_steps = 25
+ gmm.convergence_threshold = 0.00001
+ gmm.update_means = True
+ gmm.update_variances = True
+ gmm.update_weights = True
+
+ # Run ML
+ gmm.fit(ar)
+
+ # Test results
+ # Load torch3vision reference
+ meansML_ref = load_array(resource_filename("bob.learn.em", "data/meansAfterML.hdf5"))
+ variancesML_ref = load_array(resource_filename("bob.learn.em", "data/variancesAfterML.hdf5"))
+ weightsML_ref = load_array(resource_filename("bob.learn.em", "data/weightsAfterML.hdf5"))
+
+ # Compare to current results
+ np.testing.assert_allclose(gmm.means, meansML_ref, atol=3e-3)
+ np.testing.assert_allclose(gmm.variances, variancesML_ref, atol=3e-3)
+ np.testing.assert_allclose(gmm.weights, weightsML_ref, atol=1e-4)
+
+def test_gmm_MAP_dask():
+ """Test a GMMMachine for MAP with a dask array as data."""
+ ar = da.array(load_array(resource_filename("bob.learn.em", "data/dataforMAP.hdf5")))
+
+ # Initialize GMMMachine
+ n_gaussians = 5
+ prior_gmm = GMMMachine(n_gaussians)
+ prior_gmm.means = load_array(resource_filename("bob.learn.em", "data/meansAfterML.hdf5"))
+ prior_gmm.variances = load_array(resource_filename("bob.learn.em", "data/variancesAfterML.hdf5"))
+ prior_gmm.weights = load_array(resource_filename("bob.learn.em", "data/weightsAfterML.hdf5"))
+
+ threshold = 0.001
+ prior_gmm.variance_thresholds = threshold
+
+ # Initialize MAP Trainer
+ prior = 0.001
+ accuracy = 0.00001
+ gmm = GMMMachine(
+ n_gaussians,
+ trainer="map",
+ ubm=prior_gmm,
+ convergence_threshold=prior,
+ max_fitting_steps=1,
+ update_means=True,
+ update_variances=False,
+ update_weights=False,
+ mean_var_update_threshold=accuracy,
+ relevance_factor=None,
+ )
+ gmm.variance_thresholds = threshold
+
+ # Train
+ gmm = gmm.fit(ar)
+
+ # Test results
+ # Load torch3vision reference
+ meansMAP_ref = load_array(resource_filename("bob.learn.em", "data/meansAfterMAP.hdf5"))
+ variancesMAP_ref = load_array(resource_filename("bob.learn.em", "data/variancesAfterMAP.hdf5"))
+ weightsMAP_ref = load_array(resource_filename("bob.learn.em", "data/weightsAfterMAP.hdf5"))
+
+ # Compare to current results
+ # Gaps are quite large. This might be explained by the fact that there is no
+ # adaptation of a given Gaussian in torch3 when the corresponding responsibilities
+ # are below the responsibilities threshold
+ np.testing.assert_allclose(gmm.means, meansMAP_ref, atol=2e-1)
+ np.testing.assert_allclose(gmm.variances, variancesMAP_ref, atol=1e-4)
+ np.testing.assert_allclose(gmm.weights, weightsMAP_ref, atol=1e-4)
--
GitLab