Uncertainty Quantification - Interactive Visualization

Uncertainty quantification distinguishes two types of uncertainty in model predictions. Epistemic uncertainty arises from insufficient training data - the model does not know enough about a region of input space. Aleatoric uncertainty is irreducible noise inherent to the data generating process. MC dropout approximates a Bayesian neural network by running many stochastic forward passes and measuring the variance in predictions.

• See confidence intervals widen dramatically in regions with no training data - epistemic uncertainty in action
• Enable MC dropout and watch the spread of 50 stochastic predictions at each input point - the variance is your uncertainty estimate
• Understand aleatoric uncertainty: even with infinite data, some outcomes are inherently random - noise that cannot be reduced
• Compare a well-calibrated model (predicted confidence matches actual accuracy) vs an overconfident model
• Learn why uncertainty quantification is critical in medical AI, autonomous vehicles, and any high-stakes deployment

Part of the EngineersOfAI Interactive 3D - free interactive visualizations covering every major concept in machine learning and AI engineering. Hover any element for a plain-English explanation. No code required.