Compute NFVS from FVS and signed interaction graph about biobalm

[giang-trinh]

I will also add my greedy algorithm for approximating a minimum NFVS. This algorithm uses FVS_python3 to compute an FVS first. Then, it iteratively removes nodes from the FVS until there is no negative cycles. We will compare the two algorithms.

In my experience, exact methods using ILP, SAT, or ASP do not scale well for very large models in most cases. In addition, the time spent for computing a minimum NFVS may be much larger than the time spent for Preprocessing SSF on the extra states. Hence, I do not recommend using exact methods here.

[daemontus]

I have a greedy algorithm in AEON that does this. Sadly it's not exported in AEON.py yet, but it shouldn't be a problem to reimplement it on a generic DiGraph. So the algorithm is not optimal (the NFVS is not necessarily minimal), but works quite all right on most networks.

Later, I can try to add some method based on SAT to compute the actual minimal NFVS, but that might not be particularly fast for very large networks, so I'm not sure if we want that anyway... Maybe eventually, we can also compare the size of the greedy NFVS and the number of independent negative cycles in the graph and if the discrepancy is too big, we can run the SAT based detection to get the optimal result :)