I was thinking whether sklearn.preprocessing.StandardScaler() performs same as subtracting mean and dividing by standard deviation. At least thats what doc seem to say.