Aspects of Randomness in!Neural Graph Structures
*Michelle Rudolph-Lilith!Lyle E Muller
JournalClub :: Gif-sur-Yvette :: 2013/04/08
Graph-Theory Preliminaries!
number of nodes
Graph-Theory Preliminaries!
number of nodesnumber of edges
Graph-Theory Preliminaries!
number of nodesnumber of edgesadjacency matrix
1
32
4
Graph-Theory Preliminaries!
number of nodesnumber of edgesadjacency matrix
total adjacency
Graph-Theory Preliminaries!
number of nodesnumber of edgesadjacency matrix
total adjacency
connectedness
Graph-Theory Preliminaries!
number of nodesnumber of edgesadjacency matrix
total adjacency
connectedness
asymmetry index
Graph-Theory Preliminaries!
number of nodesnumber of edgesadjacency matrix
total adjacency
connectedness
asymmetry index
Graph-Theory Preliminaries!
number of nodesnumber of edgesadjacency matrix
total adjacency
connectedness
asymmetry index
undirected!graph
Graph-Theory Preliminaries!
number of nodesnumber of edgesadjacency matrix
total adjacency
connectedness
asymmetry index
undirected!graph
neural graph
C. elegans
CE1 306 2345CE2 297 2345
CE3 279 2996
CatCC1 95 2126CC2 52 818
Macaque
MB1 383 6602MC1 71 746
MC2 94 2390
MNC1 47 505
MVC1 30 311
MVC2 32 315
Historical Neural Graphs!
MB1
Adjacency, Connectedness, Asymmetry!
CE3
Adjacency, Connectedness, Asymmetry!
Adjacency, Connectedness, Asymmetry!
CE3undirected
Adjacency, Connectedness, Asymmetry!
Adjacency, Connectedness, Asymmetry!
for random graphs:
Node-Degree Distributions!
node-degrees
directed
undirected
Node-Degree Distributions!
fitting models
Node-Degree Distributions!
node in-degree node out-degree node-degree
CE3
Node-Degree Distributions!
node in-degree node out-degree node-degree
CC1
Structural Equivalence!
Euclidean distance
Pearson correlation coefficient
Structural Equivalence!
for random graphs: for random graphs:
Structural Equivalence!
correlation coefficientof node end-degrees
Structural Equivalence!
Nearest Neighbor Degrees!
average nearest neighbor degrees
directed
undirected
Nearest Neighbor Degrees!
Nearest Neighbor Degrees!
assortativity coefficient
Nearest Neighbor Degrees!
Summary and Conclusion!
node degree distributions are in accordance with a gamma model, supporting the idea of a simple local mechanism responsible for generating neural graphs
structural equivalence analysis suggests independent random distribution of node connections for different nodes, but strong correlations between in-coming and out-going connections for the same node
a weak disassortative tendency was observed, suggesting that in neural graphs nodes tend to connect with nodes of slightly higher degree
Contrary to many results reported in the neuroscientific literature, structural neural graphs show a consistency with randomness, as opposed to a consistency with more abstract models of graph construction, such the scale-free graph!