Fig. 1

Illustration of three source identification frameworks. In all three cases, the epidemic process starts from a single source, which needs to be estimated based on different observations. a In the snapshot framework, every node reveals whether it had been infected (marked in orange) by some time t, when the snapshot is taken (in this case at time \(t=6.7\), where time is measured in days, relative to some arbitrary initial time, since the infection time of the source is unknown in all cases). b In the sensor-based framework, each node is assumed to have already been infected, and a few sensor nodes (marked in orange) report their infection time \(t_i\). c In the Source Identification via Contact Tracking Framework (SICTF), the source identification task starts when the first node is hospitalized (marked in solid black), and initially most of the network is unknown to the algorithm (marked in grey). Then, in each step, the algorithm proceeds to explore the network and query the infection state of the nodes in an adaptive way, and queried nodes reveal their symptom onset time