[MRG+1]Update discriminant analysis code for better memory usage (#10904

)

doc/whats_new/v0.20.rst

@@ -157,6 +157,10 @@ Classifiers and regressors
   :class:`neighbors.RadiusNeighborsClassifier` are now
   parallelized according to ``n_jobs`` regardless of ``algorithm``.
   :issue:`8003` by :user:`Joël Billaud <recamshak>`.
+
+- Memory usage improvement for :func:`_class_means` and :func:`_class_cov`
+  in :class:`discriminant_analysis`.
+  :issue:`10898` by :user:`Nanxin Chen <bobchennan>`.`
 
 Cluster
 

sklearn/discriminant_analysis.py

@@ -83,18 +83,18 @@ def _class_means(X, y):
 
     Returns
     -------
-    means : array-like, shape (n_features,)
+    means : array-like, shape (n_classes, n_features)
         Class means.
     """
-    means = []
-    classes = np.unique(y)
-    for group in classes:
-        Xg = X[y == group, :]
-        means.append(Xg.mean(0))
-    return np.asarray(means)
+    classes, y = np.unique(y, return_inverse=True)
+    cnt = np.bincount(y)
+    means = np.zeros(shape=(len(classes), X.shape[1]))
+    np.add.at(means, y, X)
+    means /= cnt[:, None]
+    return means
 
 
-def _class_cov(X, y, priors=None, shrinkage=None):
+def _class_cov(X, y, priors, shrinkage=None):
     """Compute class covariance matrix.
 
     Parameters
@@ -120,11 +120,11 @@ def _class_cov(X, y, priors=None, shrinkage=None):
         Class covariance matrix.
     """
     classes = np.unique(y)
-    covs = []
-    for group in classes:
+    cov = np.zeros(shape=(X.shape[1], X.shape[1]))
+    for idx, group in enumerate(classes):
         Xg = X[y == group, :]
-        covs.append(np.atleast_2d(_cov(Xg, shrinkage)))
-    return np.average(covs, axis=0, weights=priors)
+        cov += priors[idx] * np.atleast_2d(_cov(Xg, shrinkage))
+    return cov
 
 
 class LinearDiscriminantAnalysis(BaseEstimator, LinearClassifierMixin,