TD or not TD: Analyzing the Role of Temporal Differencing in Deep Reinforcement Learning.
Artemij AmiranashviliAlexey DosovitskiyVladlen KoltunThomas BroxPublished in: ICLR (Poster) (2018)
Keyphrases
- reinforcement learning
- temporal difference
- reinforcement learning algorithms
- eligibility traces
- temporal difference learning
- td learning
- function approximation
- model free
- learning algorithm
- policy evaluation
- evaluation function
- state space
- monte carlo
- markov decision processes
- step size
- temporal difference methods
- action selection
- multi agent
- supervised learning
- function approximators
- reinforcement learning methods
- policy search
- machine learning
- convergence rate
- control problems
- hidden markov models
- data sets
- real time