ICMLA
2007
13 years 5 months ago
2007
Given any generative classifier based on an inexact density model, we can define a discriminative counterpart that reduces its asymptotic error rate, while increasing the estima...