[KongchangAI]
Conceptmarginal coverage

Marginal Coverage

A coverage guarantee in conformal prediction that only requires the average inclusion probability of the prediction set over all samples to be no lower than a preset level, without guaranteeing coverage for conditional subgroups.

Timeline (last 90 days)

Nov 18

Standard conformal prediction provides marginal coverage guarantees, i.e., over a large number of predictions, the proportion in which the true label falls within the prediction set is no lower than a preset level.

Unverified50%
Oct 5

Conditional coverage is stronger than marginal coverage, requiring that the target coverage rate be maintained for any given input condition, whereas marginal coverage only guarantees coverage in an expected sense across all test samples.

Unverified50%

All Facts (2)

Source Articles