Enhancing Discrete Particle Swarm Optimization for Hypergraph-Modeled Influence Maximization

A research team has published a new paper, 'Enhancing Discrete Particle Swarm Optimization for Hypergraph-Modeled Influence Maximization,' proposing a significant upgrade to a classic algorithm for a core network science problem. The work addresses a key limitation: most existing Influence Maximization (IM) methods, which aim to find the most influential nodes for information spread, rely on standard graphs that fail to capture the higher-order interactions prevalent in real-world systems like social groups or collaboration networks. By modeling these systems as hypergraphs (where edges can connect more than two nodes), the researchers create a more accurate but computationally challenging search space.

To solve this, the team enhanced the Discrete Particle Swarm Optimization (DPSO) algorithm. Their method represents a candidate set of seed nodes as a 'particle' and uses a novel two-layer local influence approximation to efficiently evaluate its spread potential. They also introduced a degree-based strategy to generate better initial solutions and incorporated local search rules to refine particle movement. Experimental results on both synthetic and real-world hypergraphs show their method outperforms existing baselines. Ablation studies confirmed the critical roles of both the local search component and the intelligent initialization strategy in achieving these results.