[Back]

H. Mahyar, P. Tulala, H. Rabiee, R. Grosu:
"Generative Adversarial Networks for Clustering Semiconductor Wafer Maps";
Poster: ML for Systems Workshop at NIPS 2018, Montreal, Canada; 2018-12-08; in: "Proc. of ML for Systems Workshop", ML for Systems, (2018).

Semiconductor manufacturing processes are characterized by a certain amount of process deviations. Automated detection of these production issues followed by an automated root cause analysis has a potential to increase the effectiveness of semiconductor production. Manufacturing defects exhibit typical patterns in measured wafer test data. Recognizing these patterns is an essential step for finding the root cause of production issues. This paper demonstrates that combining information Maximizing Generative Adversarial Network (InfoGAN) and Wasserstein GAN (WGAN) is suitable for extracting the most characteristic features from large real-world sensory wafer test data and in various aspects outperforms traditional unsupervised dimensionality reduction techniques. These features are then used in
subsequent clustering task to group wafers into clusters according to the patterns they exhibit. The main outcome of this work is a statistical model for recognizing spatial patterns given a wafer map. We experimentally evaluate the performance of the proposed approach over a real dataset.