Grand Truth graph vs. Simulated graph comparison

[nayoub]

@agrignard Save the simulation collision potential graph in a file @crisjf Compare simulated network with real network

[agrignard]

Two graphs are now created and updated in gama

graph<ML_people, ML_people> real_graph;
graph<ML_people, ML_people> simulated_graph;

[agrignard]

See https://github.com/gama-platform/gama/issues/2839 for the aggregated spatial graph

[crisjf]

@agrignard where are you storing the aggregated simulated spatial graph?

[agrignard]

@crisjf you finally have the generated graph.

Just trigger the saveGrpah to true when you want to save it, and it create a text file in results/generated_graph.txt

For now I just store

source.username, dest.username

Let me knwow if you need another format or other attribute (also easy to change from the code if need in

reflex saveCurrenGraph

[crisjf]

@agrignard this is great! But to do a better comparison we also need a weight between the agents. Do you think you can also save a simulated "collision potential" for each pair? Based, for example, on the time they spent in close proximity?

[agrignard]

right now we just have a connection, the time of the connection is not taken in account (can be good indeed) but it's not implemented yet

[crisjf]

We'll have to make do with what we have. Here's how it looks so far. Be aware that as soon as we include the information about the group each person belongs to, the two bars look very much the same.

[crisjf]

We should generate different versions of the simulated graph with different distance thresholds.

[agrignard]

It takes 3 steps ;-)

1. Set the distance
2. Run the model
3. Save the graph