Transfer Learning for Node Regression Applied to Spreading Prediction

Abstract

Understanding how information propagates in real-life complex networks yields a better understanding of dynamic processes such as misinformation or epidemic spreading. The recently introduced branch of machine learning methods for learning node representations offers many novel applications, one of them being the task of spreading prediction addressed in this paper. We explore the utility of the state-ofthe-art node representation learners when used to assess the effects of spreading from a given node, estimated via extensive simulations. Further, as many real-life networks are topologically similar, we systematically investigate whether the learned models generalize to previously unseen networks, showing that in some cases very good model transfer can be obtained. This work is one of the first to explore transferability of the learned representations for the task of node regression; we show there exist pairs of networks with similar structure between which the trained models can be transferred (zero-shot), and demonstrate their competitive performance. To our knowledge, this is one of the first attempts to evaluate the utility of zero-shot transfer for the task of node regression.

Toward Digital Twin Oriented Modeling of Complex Networked Systems and Their Dynamics: A Comprehensive Survey

“…They focus on transforming spatial information and other temporal features involved in a spreading process to well handled temporal information, where deep learning-based predictive models like Recursive Neural Networks and Convolutional Neural Networks can be employed to predict spreading process, as is well summarised by [303]. Researchers also start to employ network embedding approaches, as is mentioned in section III-B0b, to incorporate network information into the predictive systems, which involves typical examples of predicting epidemic spreading with graph neural networks [304] or using node regression based on transfer learning [305].…”

“…Propagation process dynamics, either on static networks or dynamic networks, has been extensively studied using nonmachine learning approaches [9], [292], where data-driven machine learning approaches have recently been a popular choice of incorporating more structurally and temporally complex network information [304], [305]. Dynamic networks involved in these studies only allow for topology changes [248], [249], [304], [305] which influences the result of spreading dynamics.…”

Towards Digital Twin Oriented Modelling of Complex Networked Systems and Their Dynamics: A Comprehensive Survey

“…They focus on transforming spatial information and other temporal features involved in a spreading process to well handled temporal information, where deep learning-based predictive models like Recursive Neural Networks and Convolutional Neural Networks can be employed to predict spreading pro-cess, as is well summarised by [295]. Researchers also start to employ network embedding approaches, as is mentioned in section III-B2, to incorporate network information into the predictive systems, which involves typical examples of predicting epidemic spreading with graph neural networks [296] or using node regression based on transfer learning [297].…”

“…1) Superposition of networks and processes: Propagation process dynamics, either on static networks or dynamic networks, has been extensively studied using non-machine learning approaches [8], [284], where data-driven machine learning approaches have recently been a popular choice of incorporating more structurally and temporally complex network information [296], [297]. Dynamic networks involved in these studies only allow for topology changes [240], [241], [296], [297] which influences the result of spreading dynamics.…”