Identifying networks with common organizational principles –Report on a talk by Dr Anatol Wegner.
Tuesday 23 May, 2017
This term’s Networks Cluster talk entitled Identifying networks with common organizational principles was delivered by Dr Anatol Wegner (UCL) who presented a new methodology for the study of networks.
Networks can be defined as a group of items that are connected in a systematic way. These could be anything from people to complex structures such as proteins interactions in a cell. The aim of the study of networks (Network science) is to find patterns in a network that relate to its functions. Many complex systems can be represented and studied as networks. Network comparison has become increasingly relevant today with social networks and technology development. There are many techniques for comparing networks but yet it still remains challenging to group correctly networks that are of a different size and density, but are hypothesized to have a similar structure. So a task is to find “signatures of generation mechanisms” which can be used for clustering networks.
Anatol and co-workers respond to this problem with a new network-comparison methodology called NetEmd. This is used to identify networks with common organizational principles by comparing the distributions of local network features (here mainly subgraph distributions) and defining a measure between them, a method that looks at the shape but not the size of a network. Throughout his talk, Anatol compared his NetEmd solution with various other methods (e.g. Graphlet Correlation Distance, GDDA and Netdis) currently used in a broad range of research areas, from the functional classification of chemical compounds to tracking the evolution of the Internet, to show that this new methodology is simple and successfully applicable in a wide variety of settings.
Anatol Wegner delivering his talk at Keble.