SNAP-tFDP: Massively Scalable Graph Layouts via Sparse Negative Sampling

2026-08-03

AI Summary

This work addresses the challenge of simultaneously achieving computational efficiency, low memory overhead, and clear community structure in large-scale graph layout. The authors propose a force-directed layout algorithm based on sparse negative sampling, which innovatively integrates a linearity-weighted t-distribution repulsion model with an edge-centric negative sampling mechanism. This approach enhances cluster separation while preserving local neighborhood structure, all without requiring complex auxiliary data structures. Leveraging lock-free bundled parallelism and GPU acceleration, the algorithm substantially reduces memory consumption and improves optimization efficiency. Experiments on twelve large-scale graphs demonstrate that the method reduces memory usage by 72% on average and produces high-quality layouts for graphs with up to 4 million nodes and 34 million edges in under 10 seconds, outperforming existing state-of-the-art approaches.

Abstract

Force-Directed Placement (FDP) is a widely used approach for network visualization, yet scaling it to massive graphs while preserving clear community structures remains a major computational and visual challenge. Existing approximation methods often rely on auxiliary data structures (e.g., spatial trees), which introduce substantial memory overhead; furthermore, traditional power-function-based forces frequently fail to separate dense clusters effectively. In this paper, we present a negative sampling-based algorithm that achieves O(|E|) time complexity with a low memory footprint, without requiring complex multi-level representations. In a first step, we introduce a linearly normalized degree-weighting scheme, which, combined with short-range bounded $t$-distribution forces, effectively untangles dense structures and enhances visual cluster separation. To optimize for this formulation efficiently, we introduce an edge-centric negative sampling strategy that naturally reconstructs the global degree-weighted objective. Furthermore, we design a lock-free, bundle-based parallelization scheme that leverages the sparsity of stochastic updates to achieve significant speedups while mitigating access conflicts. Comprehensive evaluations on 12 large-scale graphs demonstrate that the proposed method outperforms state-of-the-art algorithms in neighborhood preservation and cluster separation. Compared to existing baselines, our method reduces memory consumption by 72% on average and leverages simple GPU parallelism to generate a high-quality layout for a graph with 4 million nodes and 34 million edges in below 10 seconds.

Problem

• graph layout
• force-directed placement
• scalability
• community structure
• large-scale graphs

Innovation

• negative sampling
• force-directed layout
• t-distribution forces
• degree-weighting
• lock-free parallelization