Jargonautica
Maths and Statistics

Variance and Standard Deviation

Also called: Standard Deviation

How spread out a set of numbers is. Two systems with the same average accuracy can behave completely differently if one is consistent and the other swings wildly, and users notice the swinging far more than the average. Reporting a mean without a spread hides the thing people actually experience.

Technical view

Variance is the mean squared deviation from the mean; standard deviation is its square root, in the original units. In machine learning the word also names half of the bias-variance decomposition, meaning sensitivity to the particular training sample, which is what ensembling reduces. When reporting evaluation results, variance across seeds and across data splits is often larger than the differences between methods being compared.


Part of AI Terminology Index, section: Maths and Statistics.

How it connects

Variance and Standard Deviation links to 8 other terms. The faint lines are links between those terms. See it in the full graph.

OverfittingOverfittingEnsembleEnsembleConfidence IntervalConfidence IntervalCentral Limit TheoremCentral Limit Theor…Agent EvaluationAgent EvaluationRegression to the MeanRegression to the M…Expected ValueExpected ValueOutlierOutlierVariance and Standard Deviation