Spectral graph wavelet regularization and adaptive wavelet for the backward heat conduction problem

Abstract

This paper proposes a new regularization technique using spectral graph wavelet for backward heat conduction problem (BHCP) on the graph. The method uses the fourth-order compact difference for an approximation of differential operators. Meanwhile, the error estimate between the exact and wavelet regularized solution is derived. Adaptive node arrangement is obtained using spectral graph wavelet. The essence of the method is that the same operator can be used for the construction of the spectral graph wavelet and the approximation of the differential operator. The proposed method is applied to solve BHCP on graph arising from regular one-dimensional and two-dimensional mesh. Numerical results show that we can obtain a stable solution using spectral graph wavelet regularization. Moreover, spectral graph wavelet adaptivity and fourth-order compact difference scheme lead to a very efficient solution.