Learning Discrete Representations via Information Maximizing Self-Augmented Training.
Weihua HuTakeru MiyatoSeiya TokuiEiichi MatsumotoMasashi SugiyamaPublished in: ICML (2017)
Keyphrases
- multiple representations
- supervised learning
- feedforward neural networks
- prior knowledge
- background knowledge
- reinforcement learning
- learning process
- spatial information
- learning systems
- training process
- information processing
- active learning
- contextual information
- learning algorithm
- learning tasks
- learning problems
- learning scenarios
- mobile devices
- artificial neural networks
- learning speed
- feature space
- computer based training
- deep belief networks