Networks are natural representations of many real-world complex systems. Core/periphery structure is a prevalent feature of complex networks and it has attracted many studies over the past decades. Loosely speaking, a network exhibits core/periphery structure if it consists of a densely interconnected core and a sparsely connected periphery. Our research focuses on the methods to identify this core and periphery partition. Conventional core-periphery partition methods that rely on stochastic and modularity maximization models or spectral clustering approaches face several challenges, such as difficulty to extend to large-scale graphs, and inability to handle various nonlinear features in real-world networks. In contrast, deep learning techniques show an increasing capability to tackle network partition tasks. Our work aims to solve the core-periphery partition problem utilizing neural-based graph representation learning methods. A preliminary deep learning-based core-periphery partition model, CPAE, has been proposed. To examine the performance of our model, we apply our model to both synthetically generated network datasets and networks constructed from real-world data sets. The results demonstrated the effectiveness and robustness of our model.