Conditional Network Embeddings
This work focuses on a difficulty faced by all existing methods: many networks are fundamentally hard to embed due to their structural properties, e.g., non-uniform degree distribution, assortativity, (approximate) multipartiteness. To overcome this, we propose to condition embeddings on an informative prior that carries information about the network that is hard to embed. A combined representation of a prior and an embedding makes it possible to overcome the above problems.
Bo Kang, MSc, is a PhD student at the IDLab, Ghent University, Belgium. He holds an MSc degree in Computer Science from the University of Bonn, Germany. His primary interests are data mining and machine learning, and more specifically network representation learning, and dimensionality reduction algorithms. He has a website at http://bokang.io.