The posterior error estimation of an approximate solution for the integral equation φ(x)=f(x)+λ∫k(x,s)φ(s)ds is well known to be given by the formula max|e(x)|≤max|g(x)|/(1-M)for M≡|λ|max∫|k(x,s)|ds<1. Posterior error estimation was carried out by estimating the upper bound ε=max|h(x)|/(1-ν) for ν<1 regardless of the size of M. The overestimation in ε was evaluated by 2ν/(1-ν)relatively and by 2νε/(1+ν) absolutely. If we obtain a smaller parameter ν, the overestimation can be improved. Here, f(x), λ, and k(x,s) are given functions or parameters, g(x), h(x), and ν are calculable from known functions and the resolvent of the approximate kernel. In order to satisfy the condition ν<1, we easily evaluated the resolvent kernel using Bateman's theory. We evaluated this method by using some examples.
View full abstract