Introducing Machine Learning

Chapter 14. Probabilistic Methods: Naïve Bayes

A mathematical theory is not to be considered complete until you have made it so clear that you can explain it to the first man whom you meet on the street.

—David Hilbert, father of metamathematics

A trained model can give only a peremptory and assertive answer, but no probability about it. Is this a problem? As usual, the answer depends on the nature of the problem you’re trying to address with machine learning. To come to a conclusion, the question to the answer is: Would you accept a prediction that the same algorithm may score as quite unlikely?

At some point in history, the scientific community felt the need to add a probabilistic dimension to classification (and regression) problems. Hence, ...