【摘要】 求解对称对角占优线性方程组的GaBP(Gaussian Belief Propagation)迭代算法具有低计算复杂性和高并行性的特点。利用GaBP算法的这两个特点,便于处理大规模稀疏线性方程组的求解。为了进一步提高求解的迭代效率,使用经典迭代算法中的加速优化方法,给出了对应的多种GaBP迭代加速优化算法。从动态松驰因子的GaBP算法和Mann-GaBP迭代加速优化算法的实验结果表明,在相同精度下,所提出的加速优化算法比经典迭代算法和GaBP算法具有更高的并行执行效率。更多还原

【Abstract】 The GaBP(Gaussian Belief Propagation) algorithm is an iterative algorithm for symmetric diagonally dominant linear equations with low computational complexity and high parallelism.These two features of the GaBP algorithm are convenient for solving large-scale sparse linear equations.In order to further improve the iterative efficiency of the solution,this paper uses the accelerated optimization method in the classical iterative algorithm to give a corresponding multi-GaBP iterative acceleration optimization algorithm.The experimental results of GaBP algorithm and Mann-GaBP iterative acceleration optimization algorithm from dynamic relaxation factor show that the proposed acceleration optimization algorithm has higher parallel execution efficiency than the classical iterative algorithm and GaBP algorithm under the same precision.更多还原