Graph Neural Networks (GNNs) are a popular machine learning framework for solving various graph processing applications. This framework exploits both the graph topology and the feature vectors of the nodes. One of the important applications of GNN is in the semi-supervised node classification task. The accuracy of the node classification using GNN depends on (i) the number and (ii) the choice of the training nodes. In this article, we demonstrate that increasing the training nodes by selecting nodes from the same class that are spread out across non-contiguous subgraphs, can significantly improve the accuracy. We accomplish this by presenting a novel input intervention technique that can be used in conjunction with different GNN classification methods to increase the non-contiguous training nodes and, thereby, improve the accuracy. We also present an output intervention technique to identify misclassified nodes and relabel them with their potentially correct labels. We demonstrate on real-world networks that our proposed methods, both individually and collectively, significantly improve the accuracy in comparison to the baseline GNN algorithms. Both our methods are agnostic. Apart from the initial set of training nodes generated by the baseline GNN methods, our techniques do not need any other extra knowledge about the classes of the nodes. Thus, our methods are modular and can be used as pre-and post-processing steps with many of the currently available GNN methods to improve their accuracy.

In this study, we present a new hybrid model that integrates the anatomical and topological characteristics of a brain network. The aim is to effectively capture the structural and/or topological alterations that take place in networks as individuals transition from a healthy control state to the stage of Alzheimer's disease. The utilisation of a brain atlas allows for the assessment of the Euclidean distance between two specific regions of interest (ROIs) inside the brain. This distance is considered a metric of anatomical distance, providing a quantitative representation of an anatomical characteristic. Conversely, the measurement of topological similarity, which assesses a characteristic of topology, is determined by calculating the cosine distance between nodes following the embedding of the whole real brain network into a vector space of dimensionality d. The empirical findings obtained using real-brain network data indicate that the hybridization approach well captures the observed topological variations during the transition from a healthy cognitive state (HC) to Alzheimer's disease (AD).

Constant communities, i.e., groups of vertices that are always clustered together, independent of the community detection algorithm used, are necessary for reducing the inherent stochasticity of community detection results. Current methods for identifying constant communities require multiple runs of community detection algorithm(s). This process is extremely time consuming and not scalable to large networks. We propose a novel approach for finding the constant communities, by transforming the problem to a binary classification of edges. We apply the Otsu method from image thresholding to classify edges based on whether they are always within a community or not. Our algorithm does not require any explicit detection of communities and can thus scale to very large networks of the order of millions of vertices. Our results on real-world graphs show that our method is significantly faster and the constant communities produced have higher accuracy (as per F1 and NMI scores) than state-of-the-art baseline methods.