adaptive
adaptive copied to clipboard
Excessive optimism test
(original issue on GitLab)
opened by Anton Akhmerov (@anton-akhmerov) at 2017-12-22T13:15:54.380Z
If the loss is related to the error from the true value (e.g. the error of the integral in cquad),then:
- Eventually the estimated function must lie within the uncertainty from the true value
- If, like in cquad, the learner has a chance of estimating whether the error estimate is reliable, then the estimate must be within the uncertainty from the correct value, except in extraordinary circumstances.