# A Comparison of Metric Learning Loss Functions for End-To-End Speaker Verification

arXiv: https://arxiv.org/abs/2003.14021 · HTML: https://mlsanity.com/paper/2003.14021/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 31 Mar 2020.

## Built on by (most cited first, 8 of 8)

- [A Cookbook of Self-Supervised Learning](https://mlsanity.com/paper/2304.12210/) (2023)
- [Understanding Deep Contrastive Learning via Coordinate-wise Optimization](https://mlsanity.com/paper/2201.12680/) (2022)
- [MSA-ASR: Efficient Multilingual Speaker Attribution with frozen ASR Models](https://mlsanity.com/paper/2411.18152/) (2024)
- [PESTalk: Speech-Driven 3D Facial Animation with Personalized Emotional Styles](https://mlsanity.com/paper/2512.05121/) (2025)
- [Convoifilter: A case study of doing cocktail party speech recognition](https://mlsanity.com/paper/2308.11380/) (2023)
- [SimO Loss: Anchor-Free Contrastive Loss for Fine-Grained Supervised Contrastive Learning](https://mlsanity.com/paper/2410.05233/) (2024)
- [Learning spectro-temporal representations of complex sounds with parameterized neural networks](https://mlsanity.com/paper/2103.07125/) (2021)
- [Automating Feedback Analysis in Surgical Training: Detection, Categorization, and Assessment](https://mlsanity.com/paper/2412.00760/) (2024)

## Abstract (arXiv)

Despite the growing popularity of metric learning approaches, very little work has attempted to perform a fair comparison of these techniques for speaker verification. We try to fill this gap and compare several metric learning loss functions in a systematic manner on the VoxCeleb dataset. The first family of loss functions is derived from the cross entropy loss (usually used for supervised classification) and includes the congenerous cosine loss, the additive angular margin loss, and the center loss. The second family of loss functions focuses on the similarity between training samples and includes the contrastive loss and the triplet loss. We show that the additive angular margin loss function outperforms all other loss functions in the study, while learning more robust representations. Based on a combination of SincNet trainable features and the x-vector architecture, the network used in this paper brings us a step closer to a really-end-to-end speaker verification system, when combined with the additive angular margin loss, while still being competitive with the x-vector baseline. In the spirit of reproducible research, we also release open source Python code for reproducing our results, and share pretrained PyTorch models on torch.hub that can be used either directly or after fine-tuning.