I agree. Most supervised learning classifiers are derived based on the independent and identically distributed assumption for each (x,y) pair. To be more speci

I agree. Most supervised learning classifiers are derived based on the independent and identically distributed assumption for each (x,y) pair. To be more specific about the Naive Bayes assumption, the features of a data point are conditionally independent instead of simply independent. This means that given a certain label, these set of features are independent.