explainable ai for fab

**Explainable AI (XAI) for Fab** is the **application of interpretability methods to make ML predictions in semiconductor manufacturing understandable to process engineers** — providing explanations for why a model flagged a defect, predicted yield, or recommended a recipe change. **Key XAI Techniques** - **SHAP**: Shapley values quantify each feature's contribution to a prediction. - **LIME**: Local surrogate models explain individual predictions. - **Attention Maps**: Visualize which image regions drove a CNN's classification decision. - **Partial Dependence**: Show how changing one variable affects the prediction. **Why It Matters** - **Trust**: Engineers need to understand WHY a model made a decision before acting on it. - **Root Cause**: XAI reveals which process variables drove the prediction — accelerating root cause analysis. - **Validation**: Explanations expose when a model is using spurious correlations instead of physical causality. **XAI for Fab** is **making AI transparent to engineers** — providing the "why" behind every prediction so that process engineers can trust, validate, and learn from ML models.