关键词: 模拟, 重要性抽样, Gibbs抽样, 格子点, 数值积分

Abstract: To enhance the computing speed of high dimensional integral, an alternative Monte Carlo sampling algorithm is proposed in this paper.Firstly, the integral region is partitioned into net form. Secondly, grid points are sampled by using discrete Gibbs sampling method whose weighted functions are values of corresponding density function. Lastly, a new sampling sequence will be obtained by adding an uniform variable sequence to the original sequence correspondingly, and an estimation of the integral is given. The new sampling algorithm is as simple as the traditional numerical method. Simulating output showed that the new algorithm performs very well in computing speed.

Key words: simulation, important sampling, Gibbs sampling, grid point, numerical integral