imichaelnorris
commented
9 years ago That sounds like a possible way to reduce the number of parameters from 3 to 2 like we talked about on Tuesday. Every triplet with two excitatory bound to one inhibitory neuron becomes two edges in the edge-based graph. A triplet with one excitatory neuron and two inhibitory neurons has one or two edges, and a triplet with three inhibitory or three excitatory neurons has no edges.
The number of edges in a triplet is determined by the labeling of the edges, but that's the same as before. I'm wondering, though, if this can be used to reduce the graph problem so that we can use an SBM.
We discussed in class an SBM model approach for the vertices (orientation-based excitatory and inhibitory neurons) in the Bock paper. Would it be feasible to model this using edge-based graphs, e.g., using the synapses or edges from the excitatory neurons as the vertices in the SBM, blocking them based on the neuron type (inhibitory versus excitatory) on which they converge, and entering a "1" in the matrix if they converge onto an inhibitory neuron and a "0" otherwise? Actually, the 2nd step (blocking based on converging neuron type) might not even be necessary.