A new multiscale algorithm for solving second order boundary value problems (Q2189711)

A novel multiscale algorithm is proposed for solving second-order boundary value problems. The compact orthonormal multiscale basis functions are used in the kernel space using wavelets. The optimal approximate solution can be obtained from an orthonormal basis. The stability, convergence, and complexity of the algorithm are discussed. The accuracy of the solutions is improved by using two sets of multiscale basis functions. The basis function does not need integration, and the sparse matrix appearing in the calculation process of the algorithm reduces a lot the computational cost. Both effectiveness and feasibility of the algorithm are verified and compared with other methods by numerical examples. In the supplementary material \url{https://doi.org/10.1016/j.apnum.2020.05.020}, some \texttt{MATLAB} pictures and plots of the numerical test examples are given.