ajulian3
commented
9 years ago The networks created were natural networks. The shared attributes helped the algorithm determine which attributes were based on connectivity. How do you think this would be better implemented against the human connectome? Do you see any possibilities for improving the suggested algorithm? I feel the rule extraction algorithm could be fine tuned to rely on spatial information in a more organized way using the topology of the brain structure along with distance between voxels.
In the data mining paper, they say there are 3 purposes of the paper: to construct a network to correctly 'classify tuples,' 'prune the network while maintaining classification accuracy,' and 'extract symbolic rules from the network.' I wanted to make sure I understand what this means correctly: So, the purpose was to create a network (did they use natural networks or created ones?), and then find a way to classify nodes based on shared attributes, then we cut the network down some by removing some vertices (keeping the attributes for each node the same), and finally have an algorithm determine (based on connectivity?) which attributes correspond to which nodes. Is this correct?