Neural Map: Structured Memory for Deep Reinforcement Learning.
Emilio ParisottoRuslan SalakhutdinovPublished in: ICLR (Poster) (2018)
Keyphrases
- reinforcement learning
- fitted q iteration
- network architecture
- associative memory
- maximum a posteriori
- memory requirements
- state space
- neural network
- learning algorithm
- structured data
- function approximation
- memory space
- multi agent
- machine learning
- optimal control
- computing power
- bio inspired
- real world
- reinforcement learning algorithms
- limited memory
- memory size
- memory usage
- model free
- main memory
- learning rules
- biologically inspired
- policy search
- optimal policy
- nonlinear predictive control