Covariate Shift
A form of distribution shift in which the input distribution P(X) changes while the conditional distribution P(Y|X) remains unchanged.
Core Facts
Timeline (last 90 days)
Distribution shift includes three common types: covariate shift (the input distribution changes but the conditional distribution does not), label shift (the label marginal distribution changes) and concept drift (the conditional distribution changes over time).
Distribution shift refers to the statistical distribution difference between the training environment and the deployment environment, and is the direct mechanism by which spurious correlations cause performance collapse.
Dataset shift primarily includes three subtypes: covariate shift (changes in the distribution of input features while the conditional label distribution remains unchanged), prior probability shift (changes in the marginal distribution of labels), and concept drift (changes in the mapping relationship between inputs and outputs)
All Facts (3)
Distribution shift refers to the statistical distribution difference between the training environment and the deployment environment, and is the direct mechanism by which spurious correlations cause performance collapse.
75%UnverifiedDistribution shift includes three common types: covariate shift (the input distribution changes but the conditional distribution does not), label shift (the label marginal distribution changes) and concept drift (the conditional distribution changes over time).
50%UnverifiedDataset shift primarily includes three subtypes: covariate shift (changes in the distribution of input features while the conditional label distribution remains unchanged), prior probability shift (changes in the marginal distribution of labels), and concept drift (changes in the mapping relationship between inputs and outputs)
50%