85
Voted
ICMLA
2007
15 years 4 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...