I am trying to solve the first one of these exercises:
The zero-one loss is 1 if a sample is misclassified and 0 if it classified correctly. The empirical risk is the sample mean of the loss. The true risk is the expectation of the loss.
I do not know how to go about this. I could conduct a one-sided t-test:
$H_0:$ The true risk is greater or equal to 0.25
$H_1:$ The true risk is smaller than 0.25
But the p-value, I learned, is not the probability of the Null-Hypothesis, so this will not help.
I also thought that the true risk is not a random variable? Is this question solvable?