Two rare events coincided when Thomas Kipf, one of Max Welling‘s PhD students at the Amsterdam Machine Learning Lab (AMLab), defended his PhD “Deep Learning with Graph-Structured Representations” at the UvA on Thursday 23rd April.

The first notable event was that Thomas Kipf received his PhD cum laude. Such a cum laude distinction is limited to at most 5% of all PhD degrees, and in practice computer scientists are even stricter with their cum laude distinction. Furthermore, the event is also notable because it was the first PhD defence in the Amsterdam Data Science community that took place entirely online.

Cum Laude Distinction

As part of his PhD research, Thomas contributed seminal work on so-called “Graph Convolutional Networks” (GCN), a class of neural networks co-developed with Max Welling. GCNs extend the power of neural networks beyond numerical data (such as images or sounds) to relational data: networks of relations between objects. While relational data is common in many applications, until the introduction of GCNs there was no satisfactory way of handling such data in neural nets. Thomas and Max’s GCNs changed all that. Their paper alone received well over 3,000 citations since its publication four years ago in 2016. To put this number in perspective: most papers never receive more than a handful of citations, if any.