MFAN: Multi-modal Feature-enhanced Attention Networks for Rumor Detection

Abstract

Graph Convolutional Networks (GCNs) achieve an impressive performance due to the remarkable representation ability in learning the graph information. However, GCNs, when implemented on a deep network, require expensive computation power, making them difficult to be deployed on battery-powered devices. In contrast, Spiking Neural Networks (SNNs), which perform a bio-fidelity inference process, offer an energy-efficient neural architecture. In this work, we propose SpikingGCN, an end-to-end framework that aims to integrate the embedding of GCNs with the biofidelity characteristics of SNNs. The original graph data are encoded into spike trains based on the incorporation of graph convolution. We further model biological information processing by utilizing a fully connected layer combined with neuron nodes. In a wide range of scenarios (e.g., citation networks, image graph classification, and recommender systems), our experimental results show that the proposed method could gain competitive performance against state-of-the-art approaches. Furthermore, we show that SpikingGCN on a neuromorphic chip can bring a clear advantage of energy efficiency into graph data analysis, which demonstrates its great potential to construct environment-friendly machine learning models.

Heterogeneous graph neural networks with denoising for graph embeddings

“…MFAN [47] news, users, comment posts authorship, responsive propagation sign GAT [47] Capture implicit social connections.…”

“…For example, DDGCN [42] is a propagation-based method that takes snapshots of the propagation graph and the knowledge graph. Similarly, some heterogeneous social context-based methods [46], [47] also follow the idea of propagation-based methods, to model propagation patterns by GNNs, but with additional information obtained from a larger social context.…”

“…MFAN [47] captures unobserved links in the propagation graph. It models the social propagation context with graphs that contain three types of nodes: the news article node, user comment nodes, and user nodes (i.e., users who made comments).…”

“…In TriFN, news articles are connected to their publishing users as well as spreading users. Such a structure has been adapted in many followup studies [46], [47], [50], [51]. Both news articles with and without veracity labels are included in the heterogeneous graph.…”

“…Based on the graph representation of FANG, Mehta [51] enriches the edges in the graphs. Its idea is similar to MFAN [47] but from a node-level classification perspective. Since there are multiple types of nodes in the heterogeneous graph, different inference operators are used to augment the graph with different relationships beyond those initially seen.…”