Sample Correction
If a given class and feature value never occur together in the training set then the frequency-based probability estimate will be zero. This is problematic since it will wipe out all information in the other probabilities when they are multiplied. It is therefore often desirable to incorporate a small-sample correction in all probability estimates such that no probability is ever set to be exactly zero.
Read more about this topic: Naive Bayes Classifier
Famous quotes containing the words sample and/or correction:
“All that a city will ever allow you is an angle on it—an oblique, indirect sample of what it contains, or what passes through it; a point of view.”
—Peter Conrad (b. 1948)
“There are always those who are willing to surrender local self-government and turn over their affairs to some national authority in exchange for a payment of money out of the Federal Treasury. Whenever they find some abuse needs correction in their neighborhood, instead of applying the remedy themselves they seek to have a tribunal sent on from Washington to discharge their duties for them, regardless of the fact that in accepting such supervision they are bartering away their freedom.”
—Calvin Coolidge (1872–1933)