Graphical Neural Network Research Articles

Bias Assessment Approaches for Addressing User-Centered Fairness in GNN-Based Recommender Systems

In today’s technology-driven society, many decisions are made based on the results provided by machine learning algorithms. It is widely known that the models generated by such algorithms may present biases that lead to unfair decisions for some segments of the population, such as minority or marginalized groups. Hence, there is concern about the detection and mitigation of these biases, which may increase the discriminatory treatments of some demographic groups. Recommender systems, used today by millions of users, are not exempt from this drawback. The influence of these systems on so many user decisions, which in turn are taken as the basis for future recommendations, contributes to exacerbating this problem. Furthermore, there is evidence that some of the most recent and successful recommendation methods, such as those based on graphical neural networks (GNNs), are more sensitive to bias. The evaluation approaches of some of these biases, as those involving protected demographic groups, may not be suitable for recommender systems since their results are the preferences of the users and these do not necessarily have to be the same for the different groups. Other assessment metrics are aimed at evaluating biases that have no impact on the user. In this work, the suitability of different user-centered bias metrics in the context of GNN-based recommender systems are analyzed, as well as the response of recommendation methods with respect to the different types of biases to which these measures are addressed.

SpecKriging: GNN-Based Secure Cooperative Spectrum Sensing

Cooperative spectrum sensing (CSS) adopted by spectrum-sensing providers (SSPs) plays a key role for dynamic spectrum access and is essential for avoiding interference with licensed primary users (PUs). A typical SSP system consists of geographically distributed spectrum sensors which can be compromised to submit fake spectrum-sensing reports. In this paper, we propose SpecKriging, a new spatial-interpolation technique based on Inductive Graph Neural Network Kriging (IGNNK) for secure CSS. In SpecKriging, we first pretrain a graphical neural network (GNN) model with the historical sensing records of a few trusted anchor sensors. During system runtime, we use the trained model to evaluate the trustworthiness of non-anchor sensors’ data and also use them along with anchor sensors’ new data to retrain the model. SpecKriging outputs trustworthy sensor reports for spectrum-occupancy detection. To the best of our knowledge, SpecKriging is the first work that explores GNNs for trustworthy CSS and also incorporates the hardware heterogeneity of spectrum sensors. Extensive experiments confirm the high efficacy and efficiency of SpecKriging for trustworthy spectrum-occupancy detection even when malicious spectrum sensors constitute the majority.

Alzheimer's disease classification using cluster-based labelling for graph neural network on heterogeneous data.

Biomarkers for Alzheimer's disease (AD) diagnosis do not always correlate reliably with cognitive symptoms, making clinical diagnosis inconsistent. In this study, the performance of a graphical neural network (GNN) classifier based on data-driven diagnostic classes from unsupervised clustering on heterogeneous data is compared to the performance of a classifier using clinician diagnosis as an outcome. Unsupervised clustering on tau-positron emission tomography (PET) and cognitive and functional assessment data was performed. Five clusters embedded in a non-linear uniform manifold approximation and project (UMAP) space were identified. The individual clusters revealed specific feature characteristics with respect to clinical diagnosis of AD, gender, family history, age, and underlying neurological risk factors (NRFs). In particular, one cluster comprised mainly diagnosed AD cases. All cases within this cluster were re-labelled AD cases. The re-labelled cases are characterized by high cerebrospinal fluid amyloid beta (CSF Aβ) levels at a younger age, even though Aβ data was not used for clustering. A GNN model was trained using the re-labelled data with a multiclass area-under-the-curve (AUC) of 95.2%, higher than the AUC of a GNN trained on clinician diagnosis (91.7%; p=0.02). Overall, our work suggests that more objective cluster-based diagnostic labels combined with GNN classification may have value in clinical risk stratification and diagnosis of AD.

GNN-SubNet: disease subnetwork detection with explainable graph neural networks.

The tremendous success of graphical neural networks (GNNs) already had a major impact on systems biology research. For example, GNNs are currently being used for drug target recognition in protein-drug interaction networks, as well as for cancer gene discovery and more. Important aspects whose practical relevance is often underestimated are comprehensibility, interpretability and explainability. In this work, we present a novel graph-based deep learning framework for disease subnetwork detection via explainable GNNs. Each patient is represented by the topology of a protein-protein interaction (PPI) network, and the nodes are enriched with multi-omics features from gene expression and DNA methylation. In addition, we propose a modification of the GNNexplainer that provides model-wide explanations for improved disease subnetwork detection. The proposed methods and tools are implemented in the GNN-SubNet Python package, which we have made available on our GitHub for the international research community (https://github.com/pievos101/GNN-SubNet). Supplementary data are available at Bioinformatics online.

Estimate Three-Phase Distribution Line Parameters With Physics-Informed Graphical Learning Method

Accurate estimates of network parameters are essential for modeling, monitoring, and control in power distribution systems. In this paper, we develop a physics-informed graphical learning algorithm to estimate network parameters of three-phase power distribution systems. Our proposed algorithm uses only readily available smart meter data to estimate the three-phase series resistance and reactance of the primary distribution line segments. We first develop a parametric physics-based model to replace the black-box deep neural networks in the conventional graphical neural network (GNN). Then we derive the gradient of the loss function with respect to the network parameters and use stochastic gradient descent (SGD) to estimate the physical parameters. Prior knowledge of network parameters is also considered to further improve the accuracy of estimation. Comprehensive numerical study results show that our proposed algorithm yields high accuracy and outperforms existing methods.

Boundary-Aware graph Markov neural network for semiautomated object segmentation from point clouds

Due to the advantages of 3D point clouds over 2D optical images, the related researches on scene understanding in 3D point clouds have been increasingly attracting wide attention from academy and industry. However, many 3D scene understanding methods largely require abundant supervised information for training a data-driven model. The acquisition of such supervised information relies on manual annotations which are laborious and arduous. Therefore, to mitigate such manual efforts for annotating training samples, this paper studies a unified neural network to segment 3D objects out of point clouds interactively. Particularly, to improve the segmentation performance on the accurate object segmentation, the boundary information of 3D objects in point clouds are encoded as a boundary energy term in the Markov Random Field (MRF) model. Moreover, the MRF model with the boundary energy term is naturally integrated with the Graphical Neural Network (GNN) to obtain a compact representation for generating the boundary-preserved 3D objects. The proposed method is evaluated on two point clouds datasets obtained from different types of laser scanning systems, i.e. terrestrial laser scanning system and mobile laser scanning system. Comparative experiments show that the proposed method is superior and effective in 3D objects segmentation in different point-cloud scenarios.

Musical Bumps

<h3>Abstract</h3> Graphical neural networks (GNNs) offer the opportunity to provide new insights into the diagnosis, prognosis and contributing factors of Alzheimer’s disease (AD). However, previous studies using GNNs on AD did not incorporate tau PET neuroimaging features or, more importantly, have applied class labels based only on clinician diagnosis, which can be inconsistent. This study addresses these limitations by applying cluster-based labelling on heterogeneous data, including tau PET neuroimaging data, before applying GNN for classification. The data comprised sociodemographic, medical/family history, cognitive and functional assessments, apolipoprotein E (APoE) ε4 allele status, and combined tau PET neuroimaging and MRI data. We identified 5 clusters embedded in a 5-dimensional nonlinear UMAP space. Further projection onto a 3-dimensional UMAP space for visualisation, and using association rule and information gain algorithms, the individual clusters were revealed to have specific feature characteristics, specifically with respect to clinical diagnosis of AD, gender, parental history of AD, age, and underlying neurological risk factors. In particular, there was one cluster comprised mainly of clinician diagnosis of AD. We re-labelled the diagnostic classes based around this AD cluster such that AD cases occurred only within this cluster while those AD cases outside of it were re-labelled as a prodromal stage of AD. A post-hoc analysis supported the re-labelling of these cases given their similar brain tau PET levels, lying between those of the remaining AD and non-AD cases. We then trained a GNN model using the re-labelled data and compared the AD classification performance to that trained using clinician diagnostic labels. We found that GNN with re-labelled data had significantly higher average classification accuracy (p=0.011) and less variability (93.2±0.03%) than with clinician labelled data (90.06±0.04%). During model training, the cluster-labelled GNN model also converged faster than that using clinician labels. Overall, our work demonstrates that more objective cluster-based labels are viable for training GNN on heterogeneous data for diagnosing AD and cognitive decline, especially when aided by tau PET neuroimaging data.