Alert button

Learning from graphs with structural variation

Jun 29, 2018
Figure 1 for Learning from graphs with structural variation
Figure 2 for Learning from graphs with structural variation
Figure 3 for Learning from graphs with structural variation

Share this with someone who'll enjoy it:

We study the effect of structural variation in graph data on the predictive performance of graph kernels. To this end, we introduce a novel, noise-robust adaptation of the GraphHopper kernel and validate it on benchmark data, obtaining modestly improved predictive performance on a range of datasets. Next, we investigate the performance of the state-of-the-art Weisfeiler-Lehman graph kernel under increasing synthetic structural errors and find that the effect of introducing errors depends strongly on the dataset.

* Presented at the NIPS 2017 workshop "Learning on Distributions, Functions, Graphs and Groups"  

Share this with someone who'll enjoy it: