at the same time. ! = x$ , x& , … , x( : Dataset (each x) ∈ ! is a D-dimensional vector.) n Problem setting Ø Input : l dataset ! Ø Output : l Top-M outliers l For each outlier x) , a small set of features {, $ -. , , & -. , … , , / -. } explaining why the object is exceptional. (1 ≪ 3) l Weights of selected features {, $ -. , , & -. , … , , / -. }