Limits of inference in multilayer network reconstruction and embedding

Characterize the fundamental limits of inference for multilayer network structure, determining under what conditions hidden layers and mesoscale structures can be reliably recovered using generative models, statistical reconstruction techniques, and machine-learning-based embeddings, and when such recovery is provably infeasible.

Background

The authors note rapid advances in inference methods for multilayer networks, including generative models, statistical reconstruction, and machine-learning embeddings that aim to infer hidden layers and mesoscale organization.

Despite this progress, they explicitly state that understanding the limits of what can be inferred remains unresolved, indicating the need for formal characterizations of identifiability and recoverability in multilayer settings.