“…As such joint representation not only preserves topology structure, but also has the ability to embeds vertex attributes, edge attributes and other network related information as well. Thus, with the fixed-size and representative embedding vectors, conventional vector-based machine learning algorithms can be naturally introduced to solve diverse problems of analysis on the network, such as node classification [1,16], link prediction [7,9,28,33], node clustering [4,5,30], name disambiguation [25,32,34,35,37,38], and visualization [19,26], etc.…”