Method for adaptive sampling in examining an object and system thereof

Abstract

Examining an object, comprising: receiving potential defects, each associated with a location; performing first clustering of the potential defects to obtain first and second subsets, the clustering performed such that potential defects in the first subset are denser in a physical area than potential defects in the second subset; automatically assigning first validity probabilities to potential defects in the first and second subsets; selecting for review potential defects from the first and second subsets, according to a third policy, and in accordance with a strategy for combining top elements and randomly selected elements from the merged list; receiving indications for potential defects in part of the potential defect lists, subsequent to potential defects being reviewed; updating the policies in accordance with validation or classification of items in the first and second subsets; and repeating said assigning, selecting, receiving and updating with the updated policies, until a stopping criteria is observed.