Parallel Performance Analysis for CUDA-Based Co-rank Framework on Bipartite Graphs Heterogeneous Network
Fang Zheng, Han Tan, Fang Tian · 2018
The Co-rank is a ranking algorithm based on bipartite graphs of heterogeneous networks, which has been extensively studied and employed in the past decades. The Co-rank algorithm can utilize all kinds of objects in heterogeneous networks. However, the computational complexity of Co-rank algorithm limits its application on large scale datasets. Considering the efficiency of computation, we parallelize the algorithm with CUDA. We demonstrate our method via large-scale experiments across drug-target datasets and obtain excellent speedup. The results show that the optimization effect of Co-rank on the GPU platform is obvious.