Reviews: KONG: Kernels for ordered-neighborhood graphs

The paper proposes a family of kernels for graphs with ordered neighbourhoods and discrete labels. Any member of the family is obtained by generating a string-based representation of each node of a graph and a subsequent exploitation of the basic spectrum string kernel by Leslie et al. Specifically, the string associated to a node of the graph is efficiently and recursively generated via a tree traversal method that uses the neighbour ordering. The visit starts from a node and ends when the depth of the tree is h. Thus the visited subtrees are the starting features exploited by the kernel.