GLLA: A Unified Force-Directed Graph Layout Framework Supporting Local Adjustments

Force-directed graph layout methods are widely used in node-link diagram. However, applying uniform force computations to all nodes or node pairs limits their local adjustment ability, which is required to generate specific graph layouts. For example, users may want to increase the repulsive force among a group of densely clustered nodes while keeping the repulsive force among the other nodes unchanged, thereby facilitating the perception of connection structures among the group of nodes. In this study, a unified force-directed graph layout framework supporting local adjustments (GLLA) is proposed. GLLA consists of three core components: two unified parametric force expressions, a unified force model, and an improved stochastic gradient descent-based layout solver. The three components enable GLLA to reproduce the graph layout behaviors of mainstream force-directed algorithms and possess the capability to perform local adjustments on graph layouts. A performance experiment demonstrates that GLLA-implemented graph layouts are comparable to those produced by mainstream force-directed algorithms in graph structure preservation and readability while achieving notable improvements in time performance. Four real-world use cases show that GLLA can satisfy various local adjustment needs in diverse graph layout scenarios.