Abstract: | We consider worst-case identification in the ℓ2 norm. Given an unknown system h ε ℓ1 one wishes to choose bounded inputs u ε ℓ∞ such that given finitely many corrupted output measurements {y(k): 0 ≤ k}of Y = h*u + η, where η is noise, assumed small, one can construct an approximation g with g - h2 → 0 as η∞ → 0 and n → ∞. It is shown that inputs can be chosen such that to identify a sequence of length n with an ℓ2 error of O(η∞) one requires only O(n) measurements. A numerical example is inclu |