Description

In this paper, we will develop two kinds of dependency-sensitive distance metrics. The first captures the idea that if it can be shown that one feature can be (partly) predicted by another, then the predictable feature ...