A neighbour selection approach for identifying differential networks in conditional functional graphical models
Alessia Mapelli, Laura Carini, Francesca Ieva, Sara Sommariva · 2026
Read A neighbour selection approach for identifying differential networks in conditional functional graphical models by Alessia Mapelli, Laura Carini, Francesca Ieva, Sara Sommariva (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-02292.txt.
Introduction
Alessia Mapelli and three co authors present a method for mapping how brain areas interact, and how those links change from person to person. The data are EEG recordings, electric signals measured at the scalp over time. Links between areas matter for research and diagnosis, but they shift with age, mental state, and illness, so a single average map is not enough.
Their approach selects neighbors region by region inside a Gaussian functional graphical model, using regression that can take in age, clinical status, or other traits, whether continuous or grouped. The output is easy to read, since each link comes with a sign for whether it strengthens or weakens as a trait grows. Simulations and a real EEG study show clear gains over older methods.
Catalog record
| CREATOR | Alessia Mapelli, Laura Carini, Francesca Ieva, Sara Sommariva |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-02292
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A neighbour selection approach for identifying differential networks in conditional functional graphical models Alessia Mapelli, Laura Carini, Francesca Ieva, Sara Sommariva Abstract Estimation of brain functional connectivity from EEG data is of great importance both for medical research and diagnosis. It involves quantifying the conditional dependencies among the activity of different brain areas from the time-varying electric field recorded by sensors placed outside the scalp. These dependencies may vary within and across individuals and be influenced by covariates such as age, mental status, or disease severity. Motivated by this problem, we propose a novel neighbour selection approach based on functional-on-functional regression for the characterization of conditional Gaussian functional graphical models. We provide a fully automated, data-driven procedure for inferring conditional dependence structures among observed functional variables. In particular, pairwise interactions are directly identified and allowed to vary as a function of covariates, enabling covariate-specific modulation of connectivity patterns. Our proposed method accommodates an arbitrary number of continuous and discrete covariates. Moreover, unlike existing methods for direct estimation of differential graphical models, the proposed approach yields directly interpretable coefficients, allowing discrimination between covariate-induced increases and decreases in interaction strength. The methodology is evaluated through extensive simulation studies and an application to experimental EEG data. The results demonstrate clear advantages over existing approaches, including higher estimation accuracy and substantially reduced computational cost, especially in high-dimensional settings. \DOIsuffix bimj.200100000 \Volume 52 \Issue 61 \Year 2010 \pagespan 1 keywords Function-on-function regression; Gaussian functional graphical models; Graphical models with covariates; Group-based EEG connectivity; High-dimensional data. 1 Introduction Estimating brain functional connectivity from electroencephalographic (EEG) data consists in quantifying the conditional dependencies between the activity of different brain areas from the time-varying electric field recorded by sensors placed outside the scalp ( 22 ) . A large number of variables, such as age, mental status, severity of neurodegenerative disorders, may influence these dependencies both within and across individuals ( 16 ; 18 ) . Just to provide few examples, EEG functional connectivity have been shown to vary with the clinical phenotype of early-stage patients with Dementia with Lewy Bodies or Alzheimer’s disease ( 3 ; 6 ) . Similarly, few recent studies have demonstrated that differences in EEG functional connectivity can discriminate alcohol and drug abusers from healthy controls ( 9 ; 17 ) . Hence quantifying differences in functional connectivity across different conditions may allow to identify timely markers of such disorders. Despite these results, the use of EEG functional connectivity in clinical settings is still limited and the development of rigorous statistical techniques for computing robust estimation of the conditional dependency structure within the brain and its variation under different conditions is still an open issue ( 2 ; 23 ; 24 ) . Here we introduce a novel approach based on functional graphical models. Graphical models ( 11 ) have become a popular strategy for inferring conditional dependencies within the element of a (high-dimensional) random vector. They consist in undirected graphs where the nodes are the elements of the random vector and the edges encode pairwise conditional dependencies. When the random vector follows a multivariate Gaussian distribution such edges correspond to the non-zero elements of its precision matrix. Functional graphical models ( 19 ) extend the notion of graphical models to graphs whose nodes represent random functions thus allowing for characterizing conditional dependencies within multivariate random functions. In particular, Gaussian functional graphical models deal with multivariate Gaussian processes. A typical workflow for estimating (Gaussian) functional graphical models consists in: (i) deriving a finite dimensional representation of the observed functions by computing their projection scores with respect to a defined orthonormal basis (ii) estimating a precision matrix from the computed projection scores to infer the conditional dependence graph. Two main families of approaches currently exist for dealing with this second step ( 4 ) . The first one aims at directly estimating the whole precision matrix by optimizing the corresponding likelihood function with a group-lasso-type penalization that promotes sparsity across the entries of the matrix ( 19 ; 26 ) . The second family of methods consists in neighbour selection schemes where sparse functional regression is employed for separately estimating the neighbours of each node ( 29 ) . The main difference with the previous family is that these approaches infer the structure of the covariance operator one column at a time without the need of defining the whole precision matrix which may not be well-defined if the random functions lie in a infinite–dimension space or if the corresponding covariance operator is not partial separable ( 26 ) . Furthermore, since each node can be treated independently, these approaches efficiently scale to high dimensional settings involving a large number of nodes, and prior information can be exploited for example for limiting the analysis to a subset of nodes of interest as commonly done in seed-based functional connectivity studies. Most of the existing methods in both families assume that all the observed functions follow the same graphical models without accounting for the effects of external covariates. Only few recent works have focused on characterizing differences in the conditional dependence structures of different populations. When comparing the graphical models associated to two vector-valued samples, the differential graph is commonly characterized by studying the support of the difference between the corresponding precision matrices ( 30 ; 25 ; 13 ) . 28 extended this idea to the functional setting and proposed an algorithm called FuDGE (Functional Differential Graph Estimation) for directly estimating the functional differential graph without explicitly computing the functional graphical model of the two populations. However, such a method only accounts for the presence of one binary external variable. 15 proposed a method handling discrete covariate by assuming that all the subgroups in the population share a common graph structure with group-specific level of sparsity. Hierarchical penalization is then exploited to infer the subgroups precision matrices of the coefficients from the Karhunen-Loeve expansion of the functions in input. For handling both continuos and discrete external covariate, 12 introduce a novel linear operator, the functional precision operator, that extends the notion of precision matrix and its estimation to the functional setting. However, also this approach is based on operator analysis which can be computationally demanding and statistically inefficient in high-dimensional functional settings. Here we propose a neighbour selection strategy that extends the approach proposed by 29 to the estimation of conditional functional graphical models whose structure may vary with an arbitrary number of continuos and discrete external covariates. In detail, for each node we formulate the identification of its neighbours as a function-on-function (fof) regression problem modelling the interaction between the response variables, in our case the EEG time-series, and the covariates. The fof regression model is then approximated by projecting the observed functions in a node-specific basis determined via FPCA. The coefficient matrices of the resulting vector-on-vector (vov) regression model are estimated through a group-lasso-penalized least squares approach. The result is a fully automated, data–driven approach that, given a set of observed functions, characterizes their conditional dependence structure by directly identifying the pairwise interactions modulated by each covariate. We demonstrate the benefit of our proposed approach in the framework of functional EEG connectivity by comparing its performance with that of FUDGE and three related variants