Abstract:Counting (p,q)-bicliques in bipartite graphs poses a foundational challengewith broad applications, from densest subgraph discovery in algorithmicresearch to personalized content recommendation in practical scenarios. Despiteits significance, current leading (p,q)-biclique counting algorithms fallshort, particularly when faced with larger graph sizes and clique scales.Fortunately, the problem's inherent structure, allowing for the independentcounting of each biclique starting from every vertex, combined with asubstantial set intersections, makes it highly amenable to parallelization.Recent successes in GPU-accelerated algorithms across various domains motivateour exploration into harnessing the parallelism power of GPUs to efficientlyaddress the (p,q)-biclique counting challenge. We introduce GBC (GPU-basedBiclique Counting), a novel approach designed to enable efficient and scalable(p,q)-biclique counting on GPUs. To address major bottleneck arising fromredundant comparisons in set intersections (occupying an average of 90runtime), we introduce a novel data structure that hashes adjacency lists intotruncated bitmaps to enable efficient set intersection on GPUs via bit-wise ANDoperations. Our innovative hybrid DFS-BFS exploration strategy further enhancesthread utilization and effectively manages memory constraints. A composite loadbalancing strategy, integrating pre-runtime and runtime workload allocation,ensures equitable distribution among threads. Additionally, we employ vertexreordering and graph partitioning strategies for improved compactness andscalability. Experimental evaluations on eight real-life and two syntheticdatasets demonstrate that GBC outperforms state-of-the-art algorithms by asubstantial margin. In particular, GBC achieves an average speedup of 497.8x,with the largest instance achieving a remarkable 1217.7x speedup when p = q =8.

GPU-Accelerated Maximal Bicliques Mining Framework for Large E-commerce Networks

Accelerating Maximal Bicliques Enumeration with GPU on Large Scale Network

Efficient Maximal Biclique Enumeration on GPUs.

Accelerating Maximal Biclique Enumeration on GPUs

Enumeration of Billions of Maximal Bicliques in Bipartite Graphs Without Using GPUs

Accelerating the Bron-Kerbosch Algorithm for Maximal Clique Enumeration Using GPUs

Task-Oriented Genetic Activation for Large-Scale Complex Heterogeneous Graph Embedding.

Parallelizing Maximal Clique Enumeration on GPUs

AMBEA: Aggressive Maximal Biclique Enumeration in Large Bipartite Graph Computing

Accelerating Biclique Counting on GPU

A GPU Acceleration Framework for Motif and Discord Based Pattern Mining.

Scalable Mining of Maximal Quasi-Cliques: An Algorithm-System Codesign Approach

A GPU-based Graph Pattern Mining System.

G-thinker: Big Graph Mining Made Easier and Faster

MegBA: A GPU-Based Distributed Library for Large-Scale Bundle Adjustment

Mining Maximal Cliques on Dynamic Graphs Efficiently by Local Strategies

GAMMA: A Graph Pattern Mining Framework for Large Graphs on GPU.

Efficient and Scalable Graph Pattern Mining on GPUs

GPU-Accelerated Batch-Dynamic Subgraph Matching

Parallel Data Mining on Graphics Processors

Hierarchical Bipartite Graph Neural Networks: Towards Large-Scale E-commerce Applications