ABSTRACT
In this discussion of the possible tradeoff between resolution and noise, we have considered both matrix and integral-equation approaches that both assume that the quantities we are retrieving are values of a single function f evaluated at different values of the independent variable y. The importance of the Backus-Gilbert approach is that it shows explicitly that there is a tradeoff between resolution and noise amplification and resolution: it is not possible to minimize both the spread and the noise simultaneously.
