Consider an observer operating on a branchial graph

The branchial is already defining ancestral closeness

Observer coarse-graining a generic graph

In[]:=
GraphData/@Take[GraphData[40],5]
Out[]=

,
,
,
,

In[]:=
GraphData/@Take[GraphData[100],5]
Out[]=

,
,
,
,

In[]:=
GraphData/@Take[GraphData[200],UpTo[5]]
Out[]=

,

In[]:=
ResourceFunction["HypergraphToGraph"][ResourceFunction["WolframModel"][{{1,2,1},{1,3,4}}->{{4,5,4},{5,4,3},{1,2,5}},{{0,0,0},{0,0,0}},500,"FinalState"]]
Out[]=

What is coarse-graining on a graph?

Connect every node directly to its 2-step as well as 1-step neighbors......
In[]:=
CycleGraph[10,DirectedEdgesTrue]
Out[]=
In[]:=
GraphPower
,2
Out[]=
In[]:=
GraphPower
,2

Reduction method: “equate two nodes”

Simple case: merging of triangles in triangulation

Another form of coarse-graining: backtrack to ancestry; or forward-track to merging

Eventual merging allows coarse-graining to be consistent

Apply graph transformation rules, and say that nodes that evolve into “each other” after n steps are equivalent

Branchial case

Can either coarse grain by by-fiat conflating different paths at a particular time slice...

Or by taking time-bins and saying things are equivalent if they evolve into other in a certain time...

But to make this meaningful, we need to project to a consistent somewhat-time-independent branchial space

[ Fuzzy logic analog: the observer has a Gaussian kernel in continuous branchial space ]