Active learning aims to reduce annotation cost by predicting which samples are useful for a human expert to label. In the literature, most selection strategies are hand-designed, and it has become clear that there is no best active learning strategy. This has motivated research into meta-learning algorithms for “learning how to actively learn” (Konyushkova et al., 2019). In this paper, we compare this approach with margin sampling, reported in recent comparative studies as a very competitive heuristic (Yang et Loog, 2018; Pereira-Santos et al., 2019).