Rethinking Learning-Based Influence Maximization: Simple Neural Surrogates and Native Discrete Search

2026-08-08

AI Summary

This work addresses the inefficiency and suboptimal data utilization of existing learning-based influence maximization approaches, which often rely on complex neural architectures and continuous optimization. To overcome these limitations, the authors propose SIMBA, a framework that shifts the computational focus from representation learning to efficient discrete search. SIMBA leverages uniformly anchored node embeddings, a two-layer graph neural network surrogate model, and batched multi-swap simulated annealing to directly optimize discrete seed sets without requiring gradients or continuous relaxations. The method achieves substantial reductions in solution time while consistently outperforming state-of-the-art techniques in both influence spread effectiveness and data efficiency.

Abstract

Existing learning-based influence maximization frameworks rely heavily on complex neural architectures and continuous optimization over seed representations. We challenge this paradigm with SIMBA, a diffusion-model-agnostic framework pairing a lightweight neural surrogate with direct discrete search. SIMBA introduces three key components: 1) uniformly anchored node embeddings that eliminate initialization noise and encourage learning driven by graph topology and diffusion pattern, 2) a shallow two-layer graph neural network surrogate predicting final infection states, and 3) batched multi-swap simulated annealing that explores combinatorial seed space without gradients or continuous relaxation. By shifting compute from complex representation learning to effective discrete search, SIMBA drastically cuts time-to-solution while achieving superior influence spread and data efficiency. Our code is available at https://github.com/yl489/rethink-IM.

Problem

• influence maximization
• learning-based methods
• discrete optimization
• seed selection
• graph neural networks

Innovation

• influence maximization
• neural surrogate
• discrete search
• graph neural network
• simulated annealing