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

Researchers propose SIMBA, a diffusion-model-agnostic framework for influence maximization that replaces complex neural architectures with a lightweight two-layer graph neural network surrogate and direct discrete search via batched multi-swap simulated annealing. The method reduces time-to-solution while improving influence spread and data efficiency, with code available on GitHub.

arXiv:2608.08406v1 Announce Type: new 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.