sweichwald
commented
8 months ago Thank you for your interest!
Using the python interface of the current version v0.0.1, the adjustment distances can only be computed using all pairs of distinct nodes. Using rust, the distances can be computed for other sets of pairs of nodes combining functions provided in the crate. We plan on implementing a convenience wrapper for this in rust and to expose it in the python interface, too.
As there is a trade-off between allowing other (arbitrary?) sets of pairs of nodes and the potential to optimise and parallelise the implementation, our implementation may only offer two kinds of subsets:
- "specific nodes as treatment nodes" – $\lbrace (T,Y)\in \mathbf{V}\times\mathbf{V}\mid T\in\mathbf{T}, T\neq Y \rbrace$ for a given set $\mathbf{T}\subseteq\mathbf{V}$
- "specific nodes as effect nodes" – $\lbrace (T,Y)\in \mathbf{V}\times\mathbf{V}\mid Y\in\mathbf{Y}, T\neq Y \rbrace$ for a given set $\mathbf{Y}\subseteq\mathbf{V}$
Would that cover your intended use case? If not, we are very interested to hear if there are relevant scenarios where other (arbitrary?) sets of pairs of nodes are desirable, that we have overlooked :-)
Matyasch
Firstly, thank you for this great library!
The I-Specific Identification Distance given by Definition 5 in the paper inputs G_true, G_guess, and a set of node pairs S to be considered by the verifier. However, the current implementations of
*_aidfunctions only input G_true and G_guess, and compute the distance for all pairs of distinct nodes. Is it possible to compute distances for a given set of pairs using the python interface? If not, are you planning to support it in the future?