Virtual metrology based on relevant feature extraction and just-in-time learning approach

This paper focuses on virtual metrology in semiconductor manufacturing. Virtual metrology is requested to deal with the lack of physical measures by their estimations. This lack is due to measurement sampling strategies which aim to reduce the number of physical measurements which are time consuming and expensive. The idea is to estimate missing physical measurement using relevant features extracted from collected and non sampled data and using Just-in-time learning approach.