Track genomics, neuroscience and molecular biology.
Search papers, labs, and topics across Lattice.
This paper addresses the issue of oversmoothing in deep graph neural networks (GNNs) by introducing a recurrent architecture that injects independent Gaussian noise after each message-passing step. The authors prove that this stochastic approach results in hidden representations forming a geometrically ergodic Markov chain, thereby ensuring that the representations maintain diversity and do not collapse into a low-dimensional subspace. Their theoretical findings provide a positive lower bound on the expected stationary Dirichlet energy, demonstrating that persistent Gaussian perturbations effectively prevent asymptotic oversmoothing.
Injecting Gaussian noise into recurrent GNNs guarantees that node representations maintain diversity, defying the oversmoothing phenomenon that plagues traditional architectures.
Oversmoothing is a fundamental limitation of deep graph neural networks (GNNs), where repeated message passing causes node representations to become increasingly similar, eventually collapsing toward a low-dimensional subspace. This phenomenon limits the effective depth of message-passing architectures and motivates the search for mechanisms that preserve representation diversity. In this paper, we study a recurrent graph neural network in which independent Gaussian noise is injected after every propagation step and analyze the resulting architecture as a stochastic dynamical system. Under a standard global contraction assumption on the deterministic update, we prove that the hidden representations form a geometrically ergodic Markov chain admitting a unique invariant probability measure. Our main theoretical result establishes an explicit positive lower bound on the expected stationary Dirichlet energy, proportional to both the noise variance and the spectral gap of the underlying graph. Consequently, the stationary representations cannot collapse onto the constant manifold, providing a rigorous guarantee that asymptotic oversmoothing is prevented in the sense of non-vanishing Dirichlet energy. Our analysis reveals persistent stochastic perturbations as a fundamentally different mechanism for combating oversmoothing, complementing existing deterministic approaches based on residual connections, normalization, and graph rewiring. Finally, numerical experiments on both linear and nonlinear recurrent graph neural networks closely match the theoretical predictions, illustrating the emergence of a stationary distribution and the predicted dependence of the limiting Dirichlet energy on the noise intensity.