fix sgd squared_error loss name and default batch size in mlp

mislabeled/probe/_linear.py

@@ -282,7 +282,9 @@ def linearize_linear_model_sgd(estimator, X, y):
     else:
         raise NotImplementedError("lasso not implemented yet.")
 
-    linear = LinearModel(coef, intercept, loss=estimator.loss, regul=regul)
+    loss = "l2" if estimator.loss == "squared_error" else estimator.loss
+
+    linear = LinearModel(coef, intercept, loss=loss, regul=regul)
     return linear, X, y
 
 
@@ -362,11 +364,15 @@ def linearize_mlp(estimator, X, y):
         if y.ndim == 1:
             y = y.reshape(-1, 1)
 
+    batch_size = (
+        min(200, X.shape[0]) if estimator.batch_size == "auto" else estimator.batch_size
+    )
+
     linear = LinearModel(
         coef,
         intercept,
         loss=loss,
-        regul=estimator.alpha * estimator.batch_size / X.shape[0],
+        regul=estimator.alpha * batch_size / X.shape[0],
     )
 
     return linear, activation, y