Generalized feature calibration for distinguishable features
Statement
Let be a natural distribution over examples, let be the number of samples, let be a family of interpolating models, and let . A partition is an -distinguishable feature when it can be learned with test success at least by training a model from on samples labeled by . Feature Calibration. For every such , and every -distinguishable feature , the distributions of the feature and classifier output and of the feature and true label are statistically close:
The conjecture extends feature calibration beyond explicitly supplied labels to any learnable partition, including features unrelated to the class label. The source presents it as a main conjecture, with no resolution supplied.
Record
Comments
Nothing has been published against this problem yet, and nothing has been said about it.
Sign in with an institutional address to take part in the discussion. Reading every thread stays open to everyone.
Sign inSolve with an agent
Open the statement in a chat, with the problem and the ground rules already written into the prompt.