ghost
commented
9 years ago For the first part of your question, I think graph attributes are defined independently from vertex and edge attributes. If you only know the vertex is defined as either an excitatory or inhibatory cell you're not likely to be able to go backwards and say that this data came from the left eye (some random graph attribute) without further information about distribution, other descriptions. However you can look at the graph's network properties in order to try to infer a graph attribute. For example, detection of siezure based on the network pattern in several epilepsy patients. Perhaps you're confusing edge and vertex attributes with characteristics of the graph?
edunnwe1
dlee138
Given vertex and edge attributes, do we ever infer (or prove) graph attributes? Or are each of these sets of attributes defined completely independently? Related question: can edge attributes imply vertex attributes or vice versa? What if you were to have a directed graph of a neural circuit, say of the midget circuit in the retina, where you characterized edges as being either excitatory or inhibitory synapses in the presence of glutamate and you characterized vertices as being either ON or OFF [cells]. Perhaps you represent this as negative/positive weights for the edges and similarly negative or positive values for the vertices. I think, beginning with the knowledge of how cones respond to light, you could quickly see a relationship between the edge and vertex attributes in this circuit. Is this sort of relationship something that ever emerges in neuroscience graphs as an interesting result, or are graphs ever constructed explicitly with a relationship between v,e,g attributes in mind as a model that can exploit that relationship? Or would this be a sign of an ill-defined graph? If the former, and you wanted to demonstrate a relationship between attributes of one set and attributes of another, what would it take to demonstrate that relationship?