scikit-learn
diff --git a/‎doc/whats_new/v1.0.rst
Copy file name to clipboardExpand all lines: doc/whats_new/v1.0.rst
+5Lines changed: 5 additions & 0 deletions b/‎doc/whats_new/v1.0.rst
Copy file name to clipboardExpand all lines: doc/whats_new/v1.0.rst
+5Lines changed: 5 additions & 0 deletions
diff --git a/‎sklearn/linear_model/_logistic.py
Copy file name to clipboardExpand all lines: sklearn/linear_model/_logistic.py
+4-1Lines changed: 4 additions & 1 deletion b/‎sklearn/linear_model/_logistic.py
Copy file name to clipboardExpand all lines: sklearn/linear_model/_logistic.py
+4-1Lines changed: 4 additions & 1 deletion
@@ -113,6 +113,11 @@ Changelog
 :mod:`sklearn.linear_model`
 ...........................
 
+- |Efficiency| The implementation of :class:`linear_model.LogisticRegression`
+  has been optimised for dense matrices when using `solver='newton-cg'` and
+  `multi_class!='multinomial'`.
+  :pr:`19571` by :user:`Julien Jerphanion <jjerphan>`.
+
 - |Enhancement| Validate user-supplied gram matrix passed to linear models
   via the `precompute` argument. :pr:`19004` by :user:`Adam Midvidy <amidvidy>`.
 
 
@@ -233,7 +233,10 @@ def _logistic_grad_hess(w, X, y, alpha, sample_weight=None):
 
     def Hs(s):
         ret = np.empty_like(s)
-        ret[:n_features] = X.T.dot(dX.dot(s[:n_features]))
+        if sparse.issparse(X):
+            ret[:n_features] = X.T.dot(dX.dot(s[:n_features]))
+        else:
+            ret[:n_features] = np.linalg.multi_dot([X.T, dX, s[:n_features]])
         ret[:n_features] += alpha * s[:n_features]
 
         # For the fit intercept case.