论文部分内容阅读
为了加速定价利率衍生产品的Monte Carlo模拟,对远期测度下Libor市场模型中的漂移项用确定性函数近似,构造了一个与原问题高度相关的控制变量.然后将此控制变量算法移植到多核CPU和GPU的并行计算环境中,极大地提高了计算效率.针对利率上限的数值结果表明选取的控制变量十分有效且稳健,多核CPU具有线性加速效果,GPU相对于单核CPU具有很大的计算优势,控制变量和并行计算结合得到的加速效果大致是两者的乘积.结合控制变量和并行计算的方法可以为其他利率衍生产品如利率下限.互换期权的定价提供有效思路.
In order to speed up the Monte Carlo simulation of pricing interest rate derivative products, we construct a control variable that is highly correlated with the original problem by approximating the drift of the Libor market model with the deterministic function in the forward measure, and then migrate the control variable algorithm to multi-core CPU and GPU parallel computing environment, greatly improve the computational efficiency.The numerical results for the upper limit of interest rates show that the selected control variables are very effective and robust, multi-core CPU with linear acceleration, GPU has a large calculation relative to single-core CPU Advantages, control variables and parallel computing combined acceleration effect is roughly the product of the two.Combined with the control variables and parallel computing method for other interest rate derivatives such as lower interest rates. Swap options provide an effective way of thinking.