A Proposal for Reducing the Number of Trial-and-Error Searches for Deep Q-Networks Combined with Exploitation-Oriented Learning.
Naoki KodamaKazuteru MiyazakiTaku HaradaPublished in: ICMLA (2018)
Keyphrases
- learning problems
- learning process
- learning systems
- learning algorithm
- small number
- learning tasks
- real time
- deep learning
- reinforcement learning
- prior knowledge
- supervised learning
- memory requirements
- unsupervised learning
- incremental learning
- neural nets
- combining multiple
- maximum number
- early vision
- linear threshold
- computer networks
- background knowledge
- empirical studies
- knowledge acquisition
- bayesian networks
- metadata
- feature selection
- machine learning