Merge pull request #383 from oooo26/master

Update pypi workflow & improve codecov

Author: oooo26

.github/workflows/pypi.yml

@@ -31,10 +31,11 @@ jobs:
       - name: Check metadata
         run: |
           pipx run twine check dist/*
+          realpath dist/*
 
-      - uses: actions/upload-artifact@v2
+      - uses: actions/upload-artifact@v3
         with:
-          path: dist/*.tar.gz
+          path: python/dist/*
 
   build_wheels:
     name: Wheels on ${{ matrix.os }}

python/abess/__init__.py

@@ -5,7 +5,7 @@
 # @Site    :
 # @File    : __init__.py
 
-__version__ = "0.4.2"
+__version__ = "0.4.3"
 __author__ = ("Jin Zhu, Kangkang Jiang, "
               "Junhao Huang, Yanhang Zhang, "
               "Yanhang Zhang, Shiyun Lin, "

python/abess/decomposition.py

@@ -110,14 +110,14 @@ def ratio(self, X):
         """
         X = new_data_check(self, X)
         s = np.cov(X.T)
-        if len(self.coef_.shape) == 1:
-            explain = self.coef_.T.dot(s).dot(self.coef_)
-        else:
-            explain = np.sum(np.diag(self.coef_.T.dot(s).dot(self.coef_)))
-        if isinstance(s, (int, float)):
-            full = s
-        else:
-            full = np.sum(np.diag(s))
+        # if len(self.coef_.shape) == 1:
+        #     explain = self.coef_.T.dot(s).dot(self.coef_)
+        # else:
+        explain = np.sum(np.diag(self.coef_.T.dot(s).dot(self.coef_)))
+        # if isinstance(s, (int, float)):
+        #     full = s
+        # else:
+        full = np.sum(np.diag(s))
         return explain / full
 
     def fit(self, X=None, is_normal=False,
@@ -261,7 +261,7 @@ def fit(self, X=None, is_normal=False,
             elif self.screening_size > p:
                 raise ValueError(
                     "screening size should be smaller than X.shape[1].")
-            elif self.screening_size < max(support_sizes):
+            elif self.screening_size < np.nonzero(support_sizes)[0].max() + 1:
                 raise ValueError(
                     "screening size should be more than max(support_size).")
 
@@ -290,11 +290,11 @@ def fit(self, X=None, is_normal=False,
                 "number should be an positive integer and"
                 " not bigger than X.shape[1].")
 
-        # Important_search
-        if (not isinstance(self.important_search, int)
-                or self.important_search < 0):
-            raise ValueError(
-                "important_search should be a non-negative number.")
+        # # Important_search
+        # if (not isinstance(self.important_search, int)
+        #         or self.important_search < 0):
+        #     raise ValueError(
+        #         "important_search should be a non-negative number.")
 
         # A_init
         if A_init is None:
@@ -547,11 +547,11 @@ def fit(self, X, r, group=None, A_init=None):
         if self.splicing_type not in (0, 1):
             raise ValueError("splicing type should be 0 or 1.")
 
-        # Important_search
-        if (not isinstance(self.important_search, int)
-                or self.important_search < 0):
-            raise ValueError(
-                "important_search should be a non-negative number.")
+        # # Important_search
+        # if (not isinstance(self.important_search, int)
+        #         or self.important_search < 0):
+        #     raise ValueError(
+        #         "important_search should be a non-negative number.")
 
         # A_init
         if A_init is None:

python/abess/functions.py

@@ -72,9 +72,9 @@ def __call__(self, x):
             return value[0]
         return value
 
-    def __repr__(self):
-        return "StepFunction(x=%r, y=%r, a=%r, b=%r)" % (
-            self.x, self.y, self.a, self.b)
+    # def __repr__(self):
+    #     return "StepFunction(x=%r, y=%r, a=%r, b=%r)" % (
+    #         self.x, self.y, self.a, self.b)
 
 
 class BreslowEstimator:
@@ -125,25 +125,25 @@ def fit(self, linear_predictor, event, time):
                                                np.exp(- self.cum_baseline_hazard_.y))
         return self
 
-    def get_cumulative_hazard_function(self, linear_predictor):
-        r"""Predict cumulative hazard function.
-        Parameters
-        ----------
-        linear_predictor : array-like, shape = (n_samples,)
-            Linear predictor of risk: `X @ coef`.
-        Returns
-        -------
-        cum_hazard : ndarray, shape = (n_samples,)
-            Predicted cumulative hazard functions.
-        """
-        risk_score = np.exp(linear_predictor)
-        n_samples = risk_score.shape[0]
-        funcs = np.empty(n_samples, dtype=object)
-        for i in range(n_samples):
-            funcs[i] = StepFunction(x=self.cum_baseline_hazard_.x,
-                                    y=self.cum_baseline_hazard_.y,
-                                    a=risk_score[i])
-        return funcs
+    # def get_cumulative_hazard_function(self, linear_predictor):
+    #     r"""Predict cumulative hazard function.
+    #     Parameters
+    #     ----------
+    #     linear_predictor : array-like, shape = (n_samples,)
+    #         Linear predictor of risk: `X @ coef`.
+    #     Returns
+    #     -------
+    #     cum_hazard : ndarray, shape = (n_samples,)
+    #         Predicted cumulative hazard functions.
+    #     """
+    #     risk_score = np.exp(linear_predictor)
+    #     n_samples = risk_score.shape[0]
+    #     funcs = np.empty(n_samples, dtype=object)
+    #     for i in range(n_samples):
+    #         funcs[i] = StepFunction(x=self.cum_baseline_hazard_.x,
+    #                                 y=self.cum_baseline_hazard_.y,
+    #                                 a=risk_score[i])
+    #     return funcs
 
     def get_survival_function(self, linear_predictor):
         r"""Predict survival function.
@@ -164,7 +164,8 @@ def get_survival_function(self, linear_predictor):
                                     y=np.power(self.baseline_survival_.y, risk_score[i]))
         return funcs
 
-    def _compute_counts(self ,event, time, order=None):
+    @staticmethod
+    def _compute_counts(event, time, order=None):
         """Count right censored and uncensored samples at each unique time point.
 
         Parameters

python/abess/utilities.py

@@ -69,9 +69,9 @@ def categorical_to_dummy(x, classes=None):
     for i, x_i in enumerate(x):
         if x_i in classes:
             dummy_x[i, index[x_i]] = 1
-        else:
-            print(
-                "Data {} (index {}) is not in classes.".format(
-                    x_i,
-                    i))
+        # else:
+        #     print(
+        #         "Data {} (index {}) is not in classes.".format(
+        #             x_i,
+        #             i))
     return dummy_x

python/pytest/test_alg.py

@@ -176,6 +176,11 @@ def assert_reg(coef):
         # assert_fit(model1.coef_, model2.coef_)    # TODO
         assert_reg(model2.coef_)
 
+        # survival function
+        surv = model1.predict_survival_function(data.x)
+        time_points = np.quantile(data.y[:, 0], np.linspace(0, 0.6, 100))
+        surv[0](time_points)
+
     @staticmethod
     def test_poisson():
         np.random.seed(9)
@@ -379,8 +384,12 @@ def test_PCA():
 
         # ic
         for ic in ['aic', 'bic', 'ebic', 'gic']:
-            model4 = abess.SparsePCA(support_size=support_size, ic_type=ic)
-            model4.fit(X, is_normal=False)
+            model = abess.SparsePCA(support_size=support_size, ic_type=ic)
+            model.fit(X, is_normal=False)
+
+        # A_init
+        model = abess.SparsePCA(support_size=support_size)
+        model.fit(X, A_init=[0, 1, 2])
 
     @staticmethod
     def test_gamma():
@@ -440,6 +449,10 @@ def test_RPCA():
             model4 = abess.RobustPCA(support_size=s, ic_type=ic)
             model4.fit(X, r=r)
 
+        # always select
+        model5 = abess.RobustPCA(support_size=s, always_select=[1])
+        model5.fit(X, r=r)
+
     @staticmethod
     def test_ordinal():
         np.random.seed(0)

python/pytest/test_check.py

@@ -272,15 +272,27 @@ def test_base():
         else:
             assert False
 
+        try:
+            data = abess.make_glm_data(n=100, p=10, k=3, family='gaussian')
+            model = abess.LinearRegression()
+            model.fit(data.x, data.y)
+            model.score(data.x[:, 1:], data.y)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
         # lack of necessary parameter
         try:
+            model = abess.LinearRegression()
             model.fit(X=[[1]])
         except ValueError as e:
             print(e)
         else:
             assert False
 
         try:
+            model = abess.LinearRegression()
             model.fit(y=[1])
         except ValueError as e:
             print(e)
@@ -361,6 +373,14 @@ def test_pca():
         else:
             assert False
 
+        try:
+            model = abess.SparsePCA(screening_size=np.ones((100, 1)))
+            model.fit(data)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
         # screening_size
         model = abess.SparsePCA(screening_size=0)
         model.fit(data)
@@ -373,6 +393,14 @@ def test_pca():
         else:
             assert False
 
+        try:
+            model = abess.SparsePCA(screening_size=1, support_size=2)
+            model.fit(data)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
         # lack of necessary parameter
         try:
             model.fit()
@@ -444,6 +472,38 @@ def test_pca():
         else:
             assert False
 
+        try:
+            model1 = abess.SparsePCA(exchange_num=-1)
+            model1.fit([[1]])
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
+        try:
+            model1 = abess.SparsePCA(thread=-1)
+            model1.fit([[1]])
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
+        try:
+            model1 = abess.SparsePCA()
+            model.fit([[1]], A_init=[[0, 1, 2]])
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
+        try:
+            model1 = abess.SparsePCA()
+            model.fit([[1]], A_init=[-1])
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
     @staticmethod
     def test_rpca():
         model = abess.RobustPCA()
@@ -507,3 +567,43 @@ def test_rpca():
             print(e)
         else:
             assert False
+
+        try:
+            model1 = abess.RobustPCA(support_size=[100])
+            model1.fit([[1]], r=1)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
+        try:
+            model1 = abess.RobustPCA()
+            model1.fit([[1]], r=0.1)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
+        try:
+            model1 = abess.RobustPCA(exchange_num=-1)
+            model1.fit([[1]], r=1)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
+        try:
+            model1 = abess.RobustPCA(splicing_type=-1)
+            model1.fit([[1]], r=1)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False
+
+        try:
+            model1 = abess.RobustPCA(thread=-1)
+            model1.fit([[1]], r=1)
+        except ValueError as e:
+            print(e)
+        else:
+            assert False