Improving Self-Supervised Speaker Verification with Additive Margins
Introducing additive margins in the contrastive loss function improves the discriminative power of self-supervised speaker representations, leading to better speaker verification performance.