In References [11, 12], the maximum-entropy principle using Shannon entropy and a modified
entropy functional, respectively, were used. In this paper, as a unifying framework and generalization,
we adopt the Shannon–Jaynes entropy measure, Equation (21), and for consistency, the
variational problem is posed as the maximization of the entropy functional, and therefore the
dual (unconstrained) problem becomes a convex minimization problem. The parallels between
the conditions on i in Equations (2) and (3) and those on pi in a MAXENT formulation are
evident. Referring to the nodal sets shown in Figure 1, the basis function value i (x) is viewed
as the ‘probability of influence of a node i at x’. The maximum-entropy formulation is: find
/(x) ∈Rn
+ as the solution of the constrained optimization problem