CTQW-GraphSAGE: Trainabel Continuous-Time Quantum Walk On Graph

In recent years, Graph Neural Networks (GNNs) have made significant strides in various applications, demonstrating their potential in handling complex networked data. Simultaneously, quantum machine learning has emerged as a rapidly advancing and promising field, leveraging quantum computing principles to enhance machine learning models. Benefiting from the advancements in both GNNs and quantum machine learning, we propose a novel hybrid model called CTQW-GraphSAGE. This model aims to combine the strengths of classical and quantum approaches to improve performance on graph-related tasks. The model is built on the GraphSAGE framework, enhanced with quantum feature mapping and Continuous-Time Quantum Walk (CTQW). These enhancements are used to calculate aggregation weights for neighboring nodes relative to the target node, thereby integrating quantum properties into the classical model. We evaluate the proposed model on various benchmark datasets and compared our results with several baseline graph neural network methods. CTQW-GraphSAGE achieves comparable results to the classical models on most of the selected datasets on node classification tasks.