A naïve model for binary data

For each component,  j, compute its probability, pj, of being on in the training set. Model the probability of test vector alpha as the product of the probabilities of each of its components:

If component j of vector alpha is off

If component j of vector alpha is on

Binary vector alpha