Constellation Loss: Improving the efficiency of deep metric learning loss functions for optimal embedding.
Alfonso MedelaArtzai PicónPublished in: CoRR (2019)
Keyphrases
- loss function
- metric learning
- pairwise
- hinge loss
- maximum variance unfolding
- pairwise constraints
- reproducing kernel hilbert space
- semi supervised
- convex loss functions
- distance metric
- distance metric learning
- learning tasks
- support vector
- mahalanobis distance
- learning models
- nonlinear dimensionality reduction
- risk minimization
- multi task
- distance function
- kernel matrix
- feature space
- dimensionality reduction
- vector space
- semi supervised learning
- objective function
- multi class
- convex optimization
- similarity measure
- covariance matrix