• Mar 2026 - Our paper, "Robust and Interpretable Policy Learning for Manufacturing Process Parameters," by Yu-Hsin Hung (洪佑鑫), Bo-Ru Chen, and Chia-Yen Lee, is accepted for publication in the INFORMS Journal on Data Science.

• Feb 2026 - Welcome 鄭浩威、陳禹辰、林暐彧 join TrustAIM Lab

Our mission is to advance trustworthy AI for intelligent manufacturing systems, integrating statistical learning, optimization, and domain knowledge to enable reliable, interpretable, and deployable AI in real-world industrial environments.

Led by Professor Yu-Hsin Hung, the lab focuses on developing explainable AI (XAI), uncertainty-aware learning, and data-centric methodologies for manufacturing and decision-making systems. Our work bridges theory and practice, contributing novel methods in interpretable machine learning, robust policy learning, and machine unlearning, while addressing critical challenges in high-stakes industrial applications such as semiconductor manufacturing and process optimization.

Our research spans from foundational AI/ML methodology to applied manufacturing systems. Some projects emphasize theoretical advances in interpretability, uncertainty quantification, and optimization, while others integrate AI with complex industrial data—including high-frequency time series, sensor data, process parameters, and operational data—to improve system reliability, efficiency, and decision quality.

Trustworthy AI and learning

• Explainable AI (XAI)
  Develop interpretable models and explanation methods (e.g., LIME-based extensions, counterfactual explanations) to enhance transparency and trust in AI systems.

• Uncertainty-Aware Learning
  Quantify and leverage uncertainty (epistemic & aleatoric) for robust decision-making, model reliability, and risk-aware AI.

• Data-Centric AI & Robust Learning
  Design data-centric methodologies, including data quality assessment, distribution shift handling, and machine unlearning for resilient AI systems.

• Interpretable Policy Learning & Optimization
  Develop robust and interpretable decision-making frameworks (e.g., policy learning, reinforcement learning, and optimization under uncertainty).

AI for Intelligent Manufacturing

• Process/Fault Diagnostics & Root Cause Analysis
  Apply statistical learning and XAI to identify key factors affecting yield, quality, and system performance.

• Virtual Metrology & Dynamic Sampling
  Develop uncertainty-guided and dynamic sampling strategies for efficient and cost-effective metrology in manufacturing.

• Concept Drift & Adaptive Learning
  Design learning frameworks to handle dynamic environments, including gradual and abrupt distribution shifts in industrial processes.

• Industrial AI Systems & Deployment
  Bridge the gap between AI models and real-world systems by focusing on interpretability, robustness, and integration with manufacturing operations.

📄 Learn more: Publications | Google Scholar