In the fastest change detection problem, with the change point as the boundary, the sensor observations before the change point obey one probability distribution; the sensor observations after the change point obey another probability distribution. A natural thought was whether the fastest change detection problem could be modeled as a 2-class clustering problem. If possible, the distributed fastest change detection problem can be modeled as a two-type clustering problem with the observation vector of P dimension, and then it can be judged whether the monitored object occurs by observing the distance between the two cluster centers. changed.
In the fastest change detection problem, taking the change point as the boundary, the sensor observations before the change point obey a probability distribution; The observed value of the sensor after the change point follows another probability distribution. A natural idea is whether the fastest change detection problem can be modeled as a class 2 clustering problem. If so, the decentralized fastest change detection problem can be modeled as a class 2 clustering problem with p-dimension observation vector, and then whether the monitored object has changed can be judged by observing the distance between the two clustering centers.
In the fastest change detection problem, the change point is the boundary, and the observed values of sensors before the change point obey a probability distribution. The observed value of the sensor after the change obeys another probability distribution. A natural idea is whether the fastest change detection problem can be modeled as a class 2 clustering problem. If yes, then the distributed fastest change detection problem can be modeled as a class 2 clustering problem with the observation vector of P dimension, and then whether the monitored object has changed can be judged by observing the distance between the two clustering centers.