ENH implement gradient and allow y_i = 0 in Poisson datafit (#253)

Co-authored-by: Badr-MOUFAD <badr.moufad@emines.um6p.ma>

Author: mathurinm

skglm/datafits/single_task.py

@@ -431,15 +431,15 @@ def params_to_dict(self):
         return dict()
 
     def initialize(self, X, y):
-        if np.any(y <= 0):
+        if np.any(y < 0):
             raise ValueError(
-                "Target vector `y` should only take positive values " +
+                "Target vector `y` should only take positive values "
                 "when fitting a Poisson model.")
 
     def initialize_sparse(self, X_data, X_indptr, X_indices, y):
-        if np.any(y <= 0):
+        if np.any(y < 0):
             raise ValueError(
-                "Target vector `y` should only take positive values " +
+                "Target vector `y` should only take positive values "
                 "when fitting a Poisson model.")
 
     def raw_grad(self, y, Xw):
@@ -453,6 +453,9 @@ def raw_hessian(self, y, Xw):
     def value(self, y, w, Xw):
         return np.sum(np.exp(Xw) - y * Xw) / len(y)
 
+    def gradient(self, X, y, Xw):
+        return X.T @ self.raw_grad(y, Xw)
+
     def gradient_scalar(self, X, y, w, Xw, j):
         return (X[:, j] @ (np.exp(Xw) - y)) / len(y)