Applying graph neural networks
AA: One thing we touched on is the growth of industries’ interest in graph architectures and the use of graph neural networks, for example, in areas such as law enforcement and national security. You can imagine the applications there in detecting criminal networks. What advice can you give organizations, large or small, looking at leveraging graph analytics for the first time, in terms of the data that they’re collecting and storing, the tools they’re using, and the problems they are best suited to solve?
PV: One thing that is potentially a bit annoying when applying a graph representation architecture compared to, say, a convolutional neural network for images is that because the problem is not so rigid, we haven’t reached the point where we can just give you one architecture and say, “This is the gold standard for graph-structured data: you should always use this as the first approach.” Unfortunately, it...