Relaxed Earth Mover's Distances for Chain- and Tree-connected Spaces and their use as a Loss Function in Deep Learning.
Manuel MartínezMonica-Laura HauriletZiad Al-HalahMakarand TapaswiRainer StiefelhagenPublished in: CoRR (2016)
Keyphrases
- loss function
- deep learning
- pairwise
- unsupervised learning
- support vector
- unsupervised feature learning
- machine learning
- distance measure
- reproducing kernel hilbert space
- convex loss functions
- mental models
- distance function
- weakly supervised
- information extraction
- multi class
- higher order
- feature vectors
- similarity measure
- learning strategies
- feature selection