naive questions on removing edges and states about algebraicpetri.jl HOT 5 CLOSED

I believe this has been resolved with a new Catlab feature with cascading_rem_part! and cascading_rem_parts!. This will delete a part of the C-Set and cascade to remove all references to that part as well!

Here is an example:

using AlgebraicPetri
using Catlab.CategoricalAlgebra

# Define a petri net
sir_petri = PetriNet(3, ((1, 2), (2, 2)), (2, 3))

# Delete transitions 1 and 2
cascading_rem_parts!(sir_petri, :T, [1,2])

# Delete state 1
cascading_rem_part!(sir_petri, :S, 1)

from algebraicpetri.jl.

You can use rem_parts! from the Catlab.CategoricalAlgebra module to delete stuff. Comparing models with homomorphisms from Catlab is also productive. That will give you all the Petri net Homomorphisms between your two models. To compare models you probably actually want to compute a maximumal common submodel, which is a feature that we don't support quite yet.

from algebraicpetri.jl.

thanks james! ill play around with that. it seems like something like add_inputs! that could have a wrapper function around rem_parts for AlgebraicPetri.

i have been thinking about writing a library that has some syntax for defining and applying rules to graphs (or maybe categories?), which could easily act on petri nets (i think).

for example, the insert_before rule would be {{xs__, y}} -> {{xs, z}, {z, y}} which would match all inneighbors(y), plug them into a new vertex z, then connect z to y, leaving outneighbors(y) unchanged.

then i'd apply_rule(g, rule, vertex) or something to attempt a match + apply on a specific vertex, or apply_rule(g, rule) and apply on all vertices.

one extension of this that would be particularly useful is predicated rules like {{y, xs__}} -> {{y, xs/;P1}, {y, z}, {z, xs/;P2}}
which would allow me to wire based on properties of the matched vertices.

this would allow me to easily describe common model modifications that i may want (ie splitting infected to exposed and infected)

im coming from Graphs.jl, but i get the sense that catlab is the right way to define and express this.

from algebraicpetri.jl.

You should look at AlgebraicRewriting.jl. @kris-brown and I have a paper about rewriting on generic combinatorial datastructures that is based on some Graph Transformation literature.

from algebraicpetri.jl.

@anandijain you might also find https://github.com/AlgebraicJulia/Structured-Epidemic-Modeling useful, which contains the examples from the paper https://royalsocietypublishing.org/doi/10.1098/rsta.2021.0309. You can build up complex Petri nets from smaller "atomic" building blocks that represent simpler dynamics (e.g. recovery, death, infection).

from algebraicpetri.jl.