• assumes knowledge of the underlying distribution
• uses positive examples only for reestimation of parameters