Reproduce “Explainable AutoML (xAutoML) with adaptive modeling for yield enhancement in semiconductor smart manufacturing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Weihong Zhai, Xiupeng Shi, Yiik Diew Wong, Qing Han, Lisheng Chen
Researchers from Shanghai University and Nanyang Technological University developed xAutoML, a domain-specific, explainable Automated Machine Learning framework for semiconductor smart manufacturing that addresses yield enhancement and defect diagnosis. The system effectively manages high-dimensional, imbalanced data, achieving 92.89% accuracy for defect classification on the SECOM dataset while providing actionable insights.
Key insights
- Leveraging process knowledge for extensive feature engineering significantly enriches the feature space, leading to the discovery of highly predictive, previously uncaptured data patterns.
- A combined, two-stage feature selection approach (CAST) effectively identifies truly important features while mitigating redundancy and model dependency, crucial for complex industrial datasets.
- Integrating adaptive anomaly detection and focal loss in classification algorithms robustly handles extreme class imbalance and anomalies, enhancing model performance for critical defect diagnosis.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.