Abstract
Artificial neural networks have proven to be one of the most effective algorithms for learning and predicting information from data. Data come in many formats depending on the source and the nature of the phenomenon they pertain to. Numerous neural network models were proposed to handle the different data formats as efficiently as possible. One of the more abstract data format is graph structured data. We propose the Combined Graph Diffusion Embedding Network (CGDEN). The model combines two diffusion models to exploit the features of the graph nodes as well as the underlying structure of the graph. The model's performance was test in a node classification problem in semi-supervised setting, where only a fraction on the node labels were available in the training phase. Two benchmarking citation network datasets (Cora and Citeseer) were used to validate the model. The accuracy of the proposed model in node classification exceeded that of the previous models.