LISN#
- cosasi.source_inference.single_source.lisn.distance_prob(t, n, infection_rate=0.1)#
Approximates the probability of one node receiving the rumor/contagion from another node n edges away within time t.
- Parameters
t (int (optional)) – the observation timestep corresponding to I This is not actually used, but exists to match the format of other algorithms
n (int) – shortest path distance
infection_rate (float (optional)) – Inter-node infection efficiency from the original contagion process must be in [0, 1]
Notes
This function is defined in Section 3 of [1].
References
- 1
G. Nie and C. Quinn, “Localizing the Information Source in a Network” TrueFact 2019: KDD 2019 Workshop on Truth Discovery and Fact Checking: Theory and Practice, 2019
- cosasi.source_inference.single_source.lisn.lisn(I, G, t=None, infection_rate=0.1)#
Implements the algorithm from Localizing the Information Source in a Network to score all nodes in G [1].
- Parameters
I (NetworkX Graph) – The infection subgraph observed at a particular time step
G (NetworkX Graph) – The original graph the infection process was run on. I is a subgraph of G induced by infected vertices at observation time.
t (int (optional)) – the observation timestep corresponding to I
infection_rate (float (optional)) – Inter-node infection efficiency from the original contagion process must be in [0, 1]
Notes
Because the probabilities can be quite small, we report the log-score, rather than the raw score itself.
To our knowledge, this algorithm has no official name; it is referred to as “Algorithm 1” in its corresponding publication [1]. We dub it LISN, the acronym of the publication title (Localizing the Information Source in a Network).
Nodes outside the infection subgraph receive a score of negative infinity.
Examples
>>> result = cosasi.single_source.lisn(I, G)
References