and its ties to inductive inference. In References [11, 12], data approximation is viewed as
a problem in inductive inference. Pure mathematics follows the principle of deductive logic—
given a cause, many logical consequences can be readily inferred (Figure 7(a)). However, in
scientific problems, the reverse is more common: given certain effects or observations, the most
likely underlying causes are desired. This requires inductive logic (Figure 7(b)), as in ill-posed
inverse problems (e.g. heat conduction, scattering, image reconstruction) that arise in science and
engineering [45].