Network embedding assigns nodes in a network to low-dimensional representations and effectively preserves the network structure. Recently, a significant amount of progresses have been made toward this emerging network analysis paradigm. In this survey, we focus on categorizing and then reviewing the current development on network embedding methods, and point out its future research directions. We first summarize the motivation of network embedding. We discuss the classical graph embedding algorithms and their relationship with network embedding. Afterwards and primarily, we provide a comprehensive overview of a large number of network embedding methods in a systematic manner, covering the structure- and property-preserving network embedding methods, the network embedding methods with side information and the advanced information preserving network embedding methods. Moreover, several evaluation approaches for network embedding and some useful online resources, including the network data sets and softwares, are reviewed, too. Finally, we discuss the framework of exploiting these network embedding methods to build an effective system and point out some potential future directions.

Deep learning has been shown to be successful in a number of domains, ranging from acoustics, images, to natural language processing. However, applying deep learning to the ubiquitous graph data is non-trivial because of the unique characteristics of graphs. Recently, substantial research efforts have been devoted to applying deep learning methods to graphs, resulting in beneficial advances in graph analysis techniques. In this survey, we comprehensively review the different types of deep learning methods on graphs. We divide the existing methods into five categories based on their model architectures and training strategies: graph recurrent neural networks, graph convolutional networks, graph autoencoders, graph reinforcement learning, and graph adversarial methods. We then provide a comprehensive overview of these methods in a systematic manner mainly by following their development history. We also analyze the differences and compositions of different methods. Finally, we briefly outline the applications in which they have been used and discuss potential future research directions.

A miniaturized flume experiment was carried out to measure impact forces of viscous debris flow. The flow depth (7.2–11.2 cm), velocity (2.4–5.2 m/s) and impact force were recorded during the experiment. The impact process of debris flow can be divided into three phases by analyzing the variation of impact signals and flow regime. The three phases are the sudden strong impact of the debris flow head, continuous dynamic pressure of the body and slight static pressure of the tail. The variation of impact process is consistent with the change in the flow regime. The head has strong–rapid impact pressure, which is shown as a turbulent‐type flow; the body approximates to steady laminar flow. Accordingly, the process of debris flows hitting structures was simplified to a triangle shape, ignoring the pressure of the tail. In order to study the distribution of the debris flow impact force at different depths and variation of the impact process over time, the impact signals of slurry and coarse particles were separated from the original signals using wavelet analysis. The slurry's dynamic pressure signal appears to be a smooth curve, and the peak pressure is 12–34 kPa when the debris flow head hits the sensors, which is about 1.54 ± 0.36 times the continuous dynamic pressure of the debris flow body. The limit of application of the empirical parameter α in the hydraulic formula was also noted. We introduced the power function relationship of α and the Froude number of debris flows, and proposed a universal model for calculating dynamic pressure. The impact pressure of large particles has the characteristic of randomness. The mean frequency of large particles impacting the sensor is 210 ± 50–287 ± 29 times per second, and it is 336 ± 114–490 ± 69 times per second for the debris flow head, which is greater than that in the debris flow body. Peak impact pressure of particles at different flow depths is 40–160 kPa, which is 3.2 ± 1.5 times the impact pressure of the slurry at the bottom of the flow, 3.1 ± 0.9 times the flow in the middle, and 3.3 ± 0.9 times the flow at the surface. The differences in impact frequency indicate that most of the large particles concentrate in the debris flow head, and the number of particles in the debris flow head increases with height. This research supports the study of solid–liquid two phase flow mechanisms, and helps engineering design and risk assessment in debris flow prone areas. Copyright © 2015 John Wiley & Sons, Ltd.