How Wide and How Deep? Mitigating Over-squashing of GNNs via Channel Capacity Constrained Estimation
An information-theoretic view of hidden width, propagation depth, and compression in graph neural networks.
Papers on generative modelling, graph representation learning, and optimization.
An information-theoretic view of hidden width, propagation depth, and compression in graph neural networks.
Generate complete temporal windows at irregular query times through stable latent trajectories.
Adapt the mean–variance objective during noisy, budget-constrained black-box optimization.
Extended study of decoupled graph diffusion for multi-stock movement classification.
Separate representation learning and graph diffusion over automatically constructed stock graphs.
Learn complementary stock relations while preserving individual temporal representations.