LIME Explainer
Why did the model say that? LIME builds a local, human-readable explanation.
LIME (Local Interpretable Model-agnostic Explanations) answers a simple question: why did this model make this specific prediction? It does it by perturbing the input, querying the black box on each neighbor, and fitting a simple linear model weighted by proximity.
Click anywhere on the decision space to select a point. Then hit 'Explain' and watch: noisy neighbors appear around your point, colored by the black box's prediction. A dashed line shows the local linear boundary LIME fitted. The bar chart at the bottom tells you how much each feature pushed the prediction.
Try different models (circle, XOR, spiral) to see how the same algorithm explains different boundaries. Adjust the kernel width σ — a small σ means LIME only trusts very close neighbors; a large σ gives a more global (and often less faithful) explanation.