并行计算和稀疏存储在模糊积分上的应用

Application of parallel computing and sparse storage in fuzzy integral

近年来很多学者开展了模糊积分的相关研究,并将模糊积分应用于各种分类问题,而模糊测度的确定则是模糊积分计算的重点和难点。将并行计算和稀疏存储应用在模糊积分求解上,分别解决模糊积分计算中的时间复杂度和空间复杂度问题,并提出一种高效率模糊积分算法——基于并行和稀疏框架的模糊积分(parallel and sparse frame based fuzzy integral,PSFI)。实验表明,随着计算资源的增加,PSFI算法的加速比和效率下降较低。在变量存储上,PSFI算法在较多特征的数据集上对存储空间减少数千倍。最后,提出的PSFI算法相比之前提出的多重模糊积分(multiple nonlinear integral,MNI)算法,有较高的分类准确率。

Recently,many scholars have studied the fuzzy integral which is widely used in classification.A large number of studies show that computing the fuzzy measure is the key and difficult problem in fuzzy integral.This paper used parallel computing and sparse storage to solve time and space complexity in fuzzy measure computation respectively,and then put forward an algorithm named parallel and sparse frame based fuzzy integral(PSFI).Experimental results show that with the increase of computing resources,the speedup and efficiency of PSFI algorithm decrease slowly.In variable storage,PSFI algorithm reduces storage space by several thousand times on datasets with many features.Compared with the multiple nonlinear integral(MNI)algorithm proposed previously,PSFI algorithm has higher classification accuracy.