question about cross validation

[zhenlanwang]

First of all, I would like to thank you for setting up this website for reader like myself.

For "normal" validation, the aim is to estimate out of sample error for a given hypothesis, which depends on D-train.

For cross validation, the aim seems to have changed to estimate out of sample error for averaged hypothesis each trained with N-1 data points. Essentially, the estimates is about the expectation of out of sample error over all N-1 training set.

My question is how to resolve the inconsistency? Thank you.

Zhenlan,