A Comparison of Metric Learning Loss Functions for End-To-End Speaker Verification.
Juan Manuel CoriaHervé BredinSahar GhannaySophie RossetPublished in: SLSP (2020)
Keyphrases
- end to end
- loss function
- metric learning
- speaker verification
- pairwise
- pairwise constraints
- distance metric learning
- distance metric
- noisy environments
- semi supervised
- dimensionality reduction
- reproducing kernel hilbert space
- kernel matrix
- learning tasks
- mahalanobis distance
- multi task
- distance function
- multilayer perceptron
- neural network
- audio visual
- support vector
- multiscale
- semi supervised learning
- pattern recognition
- language model
- knn