Add tests

Author: antoinedemathelin

adapt/parameter_based/_regular.py

@@ -17,31 +17,6 @@
                          check_network)
 
 
-def get_zeros_network(name=None):
-    """
-    Return a tensorflow Model of two hidden layers
-    with 10 neurons each and relu activations. The
-    last layer is composed of one neuron with linear
-    activation.
-
-    Returns
-    -------
-    tensorflow Model
-    """
-    model = Sequential(name=name)
-    model.add(Flatten())
-    model.add(Dense(10, activation="relu",
-                    kernel_initializer="zeros",
-                    bias_initializer="zeros"))
-    model.add(Dense(10, activation="relu",
-                    kernel_initializer="zeros",
-                    bias_initializer="zeros"))
-    model.add(Dense(1, activation=None,
-                    kernel_initializer="zeros",
-                    bias_initializer="zeros"))
-    return model
-
-
 @make_insert_doc(supervised=True)
 class RegularTransferLR(BaseAdaptEstimator):
     """
@@ -183,20 +158,20 @@ def fit(self, Xt=None, yt=None, **fit_params):
             yt_ndim_below_one_ = True
 
         if beta_src.ndim <= 1:
-            beta_src.reshape(1, -1)
+            beta_src = beta_src.reshape(1, -1)
 
         if beta_src.shape[0] != yt.shape[1]:
             raise ValueError("The number of features of `yt`"
                              " does not match the number of coefs in 'estimator', "
-                             "expected %i, got %i"(beta_src.shape[0], yt.shape[1]))
+                             "expected %i, got %i"%(beta_src.shape[0], yt.shape[1]))
 
         if beta_src.shape[1] != Xt.shape[1]:            
             beta_shape = beta_src.shape[1]; Xt_shape = Xt.shape[1]
             if self.estimator_.fit_intercept:
                 beta_shape -= 1; Xt_shape -= 1
             raise ValueError("The number of features of `Xt`"
                              " does not match the number of coefs in 'estimator', "
-                             "expected %i, got %i"(beta_shape, Xt_shape))
+                             "expected %i, got %i"%(beta_shape, Xt_shape))
 
         beta_tgt = []
         for i in range(yt.shape[1]):

tests/test_finetuning.py

@@ -56,3 +56,9 @@ def test_finetune():
 
     assert np.abs(fine_tuned.encoder_.get_weights()[0] - model.encoder_.get_weights()[0]).sum() == 0.
     assert np.abs(fine_tuned.encoder_.get_weights()[-1] - model.encoder_.get_weights()[-1]).sum() == 0
+    
+    
+def test_finetune_pretrain():
+    model = FineTuning(encoder=encoder, task=task, pretrain=True, pretrain__epochs=2,
+                       loss="bce", optimizer="adam", random_state=0)
+    model.fit(Xs, ys, epochs=1, verbose=0)

tests/test_fmmd.py

@@ -1,6 +1,8 @@
 import numpy as np
+import tensorflow as tf
 
 from adapt.feature_based import fMMD
+from adapt.feature_based._fmmd import _get_optim_function
 
 np.random.seed(0)
 n = 50
@@ -24,4 +26,18 @@ def test_fmmd():
 
     fmmd.set_params(kernel="poly", degree=2, gamma=0.1)
     fmmd.fit_transform(Xs, Xt);
-    assert fmmd.features_scores_[-2:].sum() > 10 * fmmd.features_scores_[:-2].sum()
+    assert fmmd.features_scores_[-2:].sum() > 10 * fmmd.features_scores_[:-2].sum()
+    
+    
+def test_kernel_fct():
+    fct = _get_optim_function(Xs, Xt, kernel="linear")
+    fct(tf.identity(np.ones(6)))
+        
+    fct = _get_optim_function(Xs, Xt, kernel="rbf")
+    fct(tf.identity(np.ones(6)))
+    
+    fct = _get_optim_function(Xs, Xt, kernel="poly")
+    fct(tf.identity(np.ones(6)))
+    
+    
+    

tests/test_regular.py

@@ -2,6 +2,7 @@
 Test functions for regular module.
 """
 
+import pytest
 import numpy as np
 from sklearn.linear_model import LinearRegression, LogisticRegression
 import tensorflow as tf
@@ -81,6 +82,26 @@ def test_regularlr_multioutput():
     assert np.all(model.coef_.shape == (2, 5))
     assert np.all(model.intercept_.shape == (2,))
     assert model.score(X, y) > 0.9
+    
+    
+def test_regularlr_error():
+    np.random.seed(0)
+    Xs = np.random.randn(100, 5)
+    Xt = np.random.randn(100, 5)
+    ys = np.random.randn(100)
+    yt = np.random.randn(100)
+    lr = LinearRegression()
+    lr.fit(Xs, ys)
+    model = RegularTransferLR(lr, lambda_=1.)
+    model.fit(Xt, yt)
+    
+    with pytest.raises(ValueError) as excinfo:
+         model.fit(np.random.randn(100, 4), yt)
+    assert "expected 5, got 4" in str(excinfo.value)
+    
+    with pytest.raises(ValueError) as excinfo:
+         model.fit(Xt, np.random.randn(100, 2))
+    assert "expected 1, got 2" in str(excinfo.value)
 
 
 def test_regularlc_fit():
@@ -135,3 +156,15 @@ def test_regularnn_fit():
     assert np.sum(np.abs(network.get_weights()[0] - model.get_weights()[0])) < 0.001
     assert np.abs(model.predict(Xt) - yt_reg).sum() > 10
 
+    
+def test_regularnn_reg():
+    tf.random.set_seed(0)
+    np.random.seed(0)
+    network = _get_network()
+    network.fit(Xs, ys_reg, epochs=100, batch_size=100, verbose=0)
+    model = RegularTransferNN(network, regularizer="l1")
+    model.fit(Xt, yt_reg, epochs=100, batch_size=100, verbose=0)
+    
+    with pytest.raises(ValueError) as excinfo:
+         model = RegularTransferNN(network, regularizer="l3")
+    assert "l1' or 'l2', got, l3" in str(excinfo.value)