Module-based regularization improves Gaussian graphical models when observing noisy data
This work addresses the challenge of improving network inference in noisy data for researchers across sciences, representing an incremental advancement in regularization techniques.
The researchers tackled the problem of inferring relations from noisy correlational data using Gaussian graphical models by proposing module-based regularization to balance under- and overfitting, resulting in better recovery and inference of modular structure compared to the standard graphical lasso approach.
Inferring relations from correlational data allows researchers across the sciences to uncover complex connections between variables for insights into the underlying mechanisms. The researchers often represent inferred relations using Gaussian graphical models, requiring regularization to sparsify the models. Acknowledging that the modular structure of the inferred network is often studied, we suggest module-based regularization to balance under- and overfitting. Compared with the graphical lasso, a standard approach using the Gaussian log-likelihood for estimating the regularization strength, this approach better recovers and infers modular structure in noisy synthetic and real data. The module-based regularization technique improves the usefulness of Gaussian graphical models in the many applications where they are employed.