Upgrade machine learning packages

ceml/VERSION

@@ -1 +1 @@
-0.5.1
+0.5.2

requirements.txt

@@ -1,9 +1,9 @@
-numpy==1.16.4
+numpy==1.18.5
 scipy==1.4.1
 jax==0.2.0
 jaxlib==0.1.55
 cvxpy==1.1.0
-scikit-learn==0.23.2
+scikit-learn==0.24.0
 sklearn-lvq==1.1.1
-tensorflow==2.2.1
-torch==1.6.0
+tensorflow==2.3.2
+torch==1.7.1

tests/sklearn/test_sklearn_lvq.py

@@ -187,7 +187,7 @@ def test_lgmlvq():
     assert y_cf == 0
     assert model.predict(np.array([x_cf])) == 0
 
-    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, 0, features_whitelist=features_whitelist, regularization="l1", C=2.0, optimizer="bfgs", return_as_dict=False)
+    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, 0, features_whitelist=features_whitelist, regularization="l1", C=0.1, optimizer="bfgs", return_as_dict=False)
     assert y_cf == 0
     assert model.predict(np.array([x_cf])) == 0
 
@@ -211,7 +211,7 @@ def test_lgmlvq():
     assert model.predict(np.array([x_cf])) == 0
     assert all([True if i in features_whitelist else delta[i] == 0. for i in range(x_orig.shape[0])])
 
-    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, 0, features_whitelist=features_whitelist, regularization="l1", C=2.0, optimizer="bfgs", return_as_dict=False)
+    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, 0, features_whitelist=features_whitelist, regularization="l1", C=0.1, optimizer="bfgs", return_as_dict=False)
     assert y_cf == 0
     assert model.predict(np.array([x_cf])) == 0
     assert all([True if i in features_whitelist else delta[i] == 0. for i in range(x_orig.shape[0])])

tests/torch/test_torch_linearregression.py

@@ -54,7 +54,7 @@ def get_loss(self, y_target, pred=None):
 
     optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate, weight_decay=0.01) 
 
-    num_epochs = 3000
+    num_epochs = 30000
     for epoch in range(num_epochs):
         optimizer.zero_grad()
         outputs = model(x)
@@ -77,8 +77,8 @@ def get_loss(self, y_target, pred=None):
     y_target_done = lambda z: np.abs(z - y_target) < 6.
 
     optimizer = "bfgs"
-    optimizer_args = {"max_iter": 1000, "args": {"lr": 0.1}}
-    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, y_target=y_target, features_whitelist=features_whitelist, regularization="l2", C=10., optimizer=optimizer, optimizer_args=optimizer_args, return_as_dict=False, done=y_target_done)
+    optimizer_args = {"max_iter": 1000, "args": {"lr": 0.01}}
+    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, y_target=y_target, features_whitelist=features_whitelist, regularization="l1", C=35., optimizer=optimizer, optimizer_args=optimizer_args, return_as_dict=False, done=y_target_done)
     assert y_target_done(y_cf)
     assert y_target_done(model.predict(torch.from_numpy(np.array([x_cf], dtype=np.float32))))
 
@@ -88,15 +88,15 @@ def get_loss(self, y_target, pred=None):
     assert y_target_done(model.predict(torch.from_numpy(np.array([x_cf], dtype=np.float32))))
 
     optimizer = torch.optim.Adam
-    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, y_target=y_target, features_whitelist=features_whitelist, regularization="l2", C=10., optimizer=optimizer, optimizer_args=optimizer_args, return_as_dict=False, done=y_target_done)
+    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, y_target=y_target, features_whitelist=features_whitelist, regularization="l2", C=5., optimizer=optimizer, optimizer_args=optimizer_args, return_as_dict=False, done=y_target_done)
     assert y_target_done(y_cf)
     assert y_target_done(model.predict(torch.from_numpy(np.array([x_cf], dtype=np.float32))))
 
     features_whitelist = features_whitelist = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11]
 
     optimizer = "bfgs"
-    optimizer_args = {"max_iter": 1000, "args": {"lr": 0.1}}
-    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, y_target=y_target, features_whitelist=features_whitelist, regularization="l2", C=10., optimizer=optimizer, optimizer_args=optimizer_args, return_as_dict=False, done=y_target_done)
+    optimizer_args = {"max_iter": 5000, "args": {"lr": 0.01}}
+    x_cf, y_cf, delta = generate_counterfactual(model, x_orig, y_target=y_target, features_whitelist=features_whitelist, regularization="l1", C=33., optimizer=optimizer, optimizer_args=optimizer_args, return_as_dict=False, done=y_target_done)
     assert y_target_done(y_cf)
     assert y_target_done(model.predict(torch.from_numpy(np.array([x_cf], dtype=np.float32))))
     assert all([True if i in features_whitelist else delta[i] == 0. for i in range(x_orig.shape[0])])