Complexity of approximation with relative error criterion in worst, average, and probabilistic settings (Q1100662)

The authors study the complexity of approximating a continuous linear functional defined on a separable Banach space and equipped with a Gaussian measure. The quality of the approximation is measured by a relative error criterion. The main results are stated in Section 2. Additional facts explaining how approximations are computed, adaptive and nonadaptive information are explained in Section 3. The next three sections contain proofs for worst and average case results as well as properties of Gaussian measures needed to analyze the average case and probabilistic setting. The major technical section is Section 7. The authors prove that in the probabilistic setting the complexity is finite under a milde assumption while in the worst and average case settings the complexity is infinite. In Section 8 is indicated how some of the results to generalize for linear operators. For linear operators see ibid. 3, 57- 80 (1987).