Experimenting with additive margins for contrastive self-supervised speaker verification

Abstract