In the development of intelligent computing system, efficient processing of high-dimensional sparse matrix has become a key challenge to improve system performance. This article focuses on this, aiming to explore a more effective parallelization strategy for high-dimensional sparse matrix processing. By deeply analyzing the high-dimensional sparse matrix and the basic theory of parallel computing, this article proposes an adaptive parallel algorithm model based on sparse pattern recognition. Firstly, the sparse matrix pattern is identified, and regions are divided according to features. Then, the computational resources are adaptively allocated, and the computational process is optimized by using message passing communication mechanism, precompution and caching mechanism. In the specific high-performance server environment, the experiment is carried out by selecting several high-dimensional sparse matrix data sets with different characteristics. The results show that, compared with the parallel algorithm based on block partition and row partition, the calculation time of this algorithm is reduced by 30%-50% on average, the acceleration is more stable and efficient than the growth, and the average resource utilization rate is over 85%, which has obvious advantages in scalability testing. This shows that the algorithm in this article has excellent performance in dealing with high-dimensional sparse matrices and can improve the processing capacity of intelligent computing systems.
Wenyue Ma (Sun,) studied this question.