I'm looking at the general class of graph algorithms that are solvable using neural networks.
For example - https://medium.com/octavian-ai/finding-shortest-paths-with-graph-networks-807c5bbfc9c8 - refers to an interesting solution to the shortest problem using graph neural networks.
All the examples in NSL illustrate examples where the graph adds to information in an existing model. But can NSL solve graph problems itself ?
Although NSL could potentially be used to improve the accuracy or robustness of a neural network specifically trained to solve the shortest-path problem (like the RNN described in the Medium article you linked to), NSL was not developed to solve graph problems directly. That said, we'd certainly be interested in hearing about any progress in using our libraries to do so.