C3_W2_RecSysNN_Assignment why using different scalers?

Course Q&A Machine Learning Specialization Unsupervised Learning, Recommenders, Reinforcement
1

If I am right, by using StandardScaler for training data in cell #6, guarantees zero-mean values an variance equal to 1. While minmaxScaler allow to fix extremes to -1 and 1 for the values of the targets y. Why making different scaling choices?

1 Like
2

Hi @lucapug
The main difference in the scaling techniques is how to deal with :

  • distribution
  • Outliers
    for example minmaxScaler this Scaler responds well if the standard deviation is small and when a distribution is not Gaussian . This Scaler is sensitive to outliers .

you can check this articale it’s very useful

Best Regards,
Abdelrahman

3

Thank you. I will read the reference you pointed out!