Anchor Semiconductor›Anchor Semiconductor› offers Pattern Centric Machine Learning

Pattern Centric Machine Learning bridges Anchor Semiconductor's Printed Pattern Database and Design Decomposition Database to extend yield learning beyond the relatively small number of patterns captured by SEM review. It uses deviations between reference design and printed silicon to automatically label training examples, then applies supervised machine learning to predict risk for patterns that have not yet been imaged.

As new SEM images are continuously added to the printed pattern database, the model retrains and improves its prediction accuracy. The resulting risk scores are applied to the much larger design decomposition database, giving fabs broader coverage than pure statistical or geometrical signatures alone.