I think I am still a bit unclear about deterministic noise. Doesn't the amount of noise (deterministic or stochastic) depend on both the bias and variance of the noise? For a given N, the more complex noise will have a higher variance but lower bias. Hence doesn't the amount of noise depend upon N and the relative complexity difference between the target function and the hypothesis (the level of deterministic noise if you will)?
