PhD Defense: "Tensor Decomposition for Concept Recognition – an Automated and Robust Concept Recognition Software Based on GANs and Tensors"

Ahmed Wahba

September 9th (Monday), 10:00am
Harold Frank Hall (HFH), Rm 4164 (ECE Conf. Rm.)

In production test data analytics, it is often that an analysis involves the recognition of a conceptual pattern on a wafer map. A wafer pattern may hint a particular issue in the production by itself or guide the analysis into a certain direction.

In this thesis, we show how Generative Adversarial Networks (GANs) can be used to build concept recognizers. We also introduce a novel concept recognition approach based on Tensor decomposition.

Tensor-based concept recognizers are combined with GANs-based recognizers to prevent escapes, also known as adversarial examples.

The focus of this work is on automating the concept rcognition process as well as ensuring robustness. Two tensor methods are introduced, the first tensor method is based on clustering and is used to automate concept extraction. The second tensor method is used to continuously check the GANs-based recognizer’s performance to ensure robustness. The robust automated software is applied to production wafers from multiple automotive product lines.

About Ahmed Wahba:

Ahmed Wahba is a PhD. candidate in Computer Engineering advised by professor Li-C. Wang. Ahmed received his Bachelor's degree in Electrical Engineering and Master's degree in Computer Engineering from Cairo University in 2011 and 2014 respectively. His research interests are Machine Learning, Data Analytics, and Design Automation. His Doctoral research is focused on using Tensor Decomposition for Machine Learning and Data Analytics. He also has industry and research experience in Computer Architecture, RF Design, and Design Verification.

Hosted by: Li-C. Wang