Advances in network visualisation with an application to serious games
PhD thesis, Imperial College London (2021)
This thesis concerns the visualisation of networks, through an in-depth study into the node-link diagram representation. Three subtopics are explored within this space. The first is the problem of node layout, where the optimisation of a popular energy function, known as stress, is improved through an algorithm known as stochastic gradient descent. The second is the method of edge bundling, where the idea of hierarchical edge bundling is explored in the absence of a known ground truth hierarchy. Its similarity to a topologically lossless bundling method known as power-confluent drawing is then leveraged, in order to improve technical problems with the underlying algorithms. The final topic is an engineering application in the form of a serious game called EcoBuilder, which utilises the node-link diagram to visualise the dynamical behaviour of food webs. Its purpose is to crowdsource research through a citizen science approach, with outcomes in both visualisation and mathematical ecology.
Graph layout using stochastic gradient descent.
Educational game about building ecosystems.
Package for power confluent graph drawings.