A0763
Title: Zigzag filtration curve based supra-hodge convolution networks for dynamic ethereum token networks forecasting
Authors: Yuzhou Chen - Temple University (United States) [presenting]
Abstract: Graph neural networks (GNNs) offer a new powerful alternative for multivariate time series forecasting, demonstrating remarkable success in various spatiotemporal applications, from urban flow monitoring systems to health care informatics to financial analytics. Yet, such GNN models pre-dominantly capture only lower order interactions, that is, pairwise relations among nodes, and also largely ignore intrinsic time-conditioned information on the underlying topology of multivariate time series. To address these limitations, a new time-aware GNN architecture which amplifies the power of the recently emerged simplicial neural networks with a time-conditioned topological knowledge representation in the form of zigzag persistence, is proposed. That is, the new approach, Zigzag Filtration Curve based Supra-Hodge Convolution Networks (ZFC-SHCN), is built upon the two main components -- a new highly computationally efficient zigzag persistence curve and a new temporal multiplex graph representation module for learning higher-order network interactions. Theoretical properties of the proposed time-conditioned topological knowledge representation and extensive validate the new time-aware ZFC-SHCN model in conjunction with time series forecasting on Ethereum blockchain datasets are discussed. The experiments demonstrate that the ZFC-SHCN achieves state-of-the-art performance with lower requirements on computational costs.
