Fast Iterative Methods for Computing the Harmonic Information Potentials in Wireless Sensor Networks.

This paper presents a study on the dissemination of information from the source node by establishing information potentials that uses the information gradient descent method. The information gradient is computed by solving the discrete approximation of a partial difference equation. The solution to this equation is the harmonic functions which are also known as harmonic information potentials. Commonly used iterative method for solving partial difference equation to obtain the harmonic information potentials was the classical Jacobi that was found to be too slow when it dealt with large domain. This study proposes faster iterative methods using Gauss-Seidel and Successive Overrelaxation with Red-Black ordering strategy (GS-RB and SOR-RB), as well as Modified Successive Overrelaxation (MSOR) schemes. The experimental results show that the execution time of GS-RB, SOR-RB and MSOR are faster than the existing methods. The advantage of the proposed methods in terms of computational speed is clearly shown with increasingly large domain size. [ABSTRACT FROM AUTHOR]