A kernel method for unsupervised structured network inference

Network inference is the problem of infer- ring edges between a set of real-world ob- jects, for instance, interactions between pairs of proteins in bioinformatics. Current kernel- based approaches to this problem share a set of common features: (i) they are supervised and hence require labeled training data; (ii) edges in the network are treated as mutu- ally independent and hence topological prop- erties are largely ignored; (iii) they lack a sta- tistical interpretation. We argue that these common assumptions are often undesirable for network inference, and propose (i) an un- supervised kernel method (ii) that takes the global structure of the network into account and (iii) is statistically motivated. We show that our approach can explain commonly used heuristics in statistical terms. In exper- iments on social networks, dierent variants of our method demonstrate appealing predic- tive performance.