#aleatoric-uncertainty

Question 1: Single choice

easyAI-checkedProbabilistic & Graphical Models

A Gaussian process regression model is fit to noisy measurements $y = f(x) + \epsilon$ . Which statement correctly separates epistemic uncertainty from observation (aleatoric) uncertainty?

Epistemic uncertainty is about the unknown $f$ and shrinks with more data; the noise does not1
Epistemic uncertainty comes from the measurement noise, so it is the part that more data cannot remove2
Observation uncertainty grows away from the training inputs, while epistemic uncertainty is the same everywhere3
Both kinds shrink to zero as the dataset grows, only at different rates4

#aleatoric-uncertainty

Timed test