Lagrangian Duality for Constrained Deep Learning.
Ferdinando FiorettoPascal Van HentenryckTerrence W. K. MakCuong TranFederico BaldoMichele LombardiPublished in: ECML/PKDD (5) (2020)
Keyphrases
- deep learning
- unsupervised learning
- duality gap
- machine learning
- lagrange multipliers
- unsupervised feature learning
- linear programming
- mental models
- weakly supervised
- restricted boltzmann machine
- deep architectures
- pairwise
- learning algorithm
- multiple images
- named entities
- multi class
- deep belief networks
- information retrieval