Work

Real Additive Margin Softmax for Speaker Verification

Year: 2021

Type: preprint

Abstract: The additive margin softmax (AM-Softmax) loss has delivered remarkable performance in speaker verification. A supposed behavior of AM-Softmax is that it can shrink within-class variation by putting emphasis on target logits, which in turn improves margin between target and non-target classes. In this paper, we conduct a careful analysis on the behavior of AM-Softmax loss, and show that this loss does not implement real max-margin training. Based on this observation, we present a Real AM-Softmax loss which involves a true margin function in the softmax training. Experiments conducted on VoxCeleb1, SITW and CNCeleb demonstrated that the corrected AM-Softmax loss consistently outperforms the original one. The code has been released at https://gitlab.com/csltstu/sunine.

Source: arXiv (Cornell University)

DOI: 10.48550/arxiv.2110.09116

Authors: Li, Lantian, Nai, Ruiqian, Wang, Dong

Institutions: Beijing National Research Center for Information Science and Technology, Tsinghua University

Language: English

Topic: Speech Recognition and Synthesis

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funder: National Natural Science Foundation of China

Awards: G1598214476, G4491952354

OpenAlex ID: W4303493643

Cites:

Cited by: 2

Related to: 10