Loss rate estimation with incomplete data set

Loss tomography has received considerable interest in recent years. Although a number of estimators have been proposed for the tree topology, most of them do not consider data missing. To correct this, we in this paper classify data into five classes and propose four estimators, one for a type of data with missing. The estimators are proved to be the maximum likelihood ones. The work is further extended into the general topology that has hardly been explored previously, where a structure between data and models is established.