Document worth reading: “A Survey on Graph Kernels”
Graph kernels have change right into a longtime and widely-used method for fixing classification duties on graphs. This survey presents an entire overview of strategies for kernel-based graph classification developed so far 15 years. We describe and categorize graph kernels based totally on properties inherent to their design, equal to the character of their extracted graph choices, their methodology of computation and their applicability to points in apply. In an in depth experimental evaluation, we look at the classification accuracy of a giant suite of graph kernels on established benchmarks along with new datasets. We consider the effectivity of well-liked kernels with numerous baseline methods and look at the influence of creating use of a Gaussian RBF kernel to the metric induced by a graph kernel. In doing so, we uncover that simple baselines develop into aggressive after this transformation on some datasets. Moreover, we look at the extent to which current graph kernels agree of their predictions (and prediction errors) and procure a data-driven categorization of kernels as finish consequence. Finally, based totally on our experimental outcomes, we derive a practitioner’s data to kernel-based graph classification. A Survey on Graph Kernels