ReinforcementLearning performs model-free reinforcement learning in R. This implementation enables the learning of an optimal policy based on sample sequences consisting of states, actions and rewards. In addition, it supplies multiple predefined reinforcement learning algorithms, such as experience replay.
The most important functions of ReinforcementLearning are:
You can easily install the latest version of ReinforcementLearning with
# Recommended option: download and install latest version from CRAN install.packages("ReinforcementLearning") # Alternatively, install the development version from GitHub: # install.packages("devtools") devtools::install_github("nproellochs/ReinforcementLearning")
This section shows the basic functionality of how to perform reinforcement learning. First, load the corresponding package ReinforcementLearning.
The following example shows how to learn a reinforcement learning agent using input data in the form of sample sequences consisting of states, actions and rewards. The result of the learning process is a state-action table and an optimal policy that defines the best possible action in each state.
# Generate sample experience in the form of state transition tuples data <- sampleGridSequence(N = 1000) head(data) #> State Action Reward NextState #> 1 s4 left -1 s4 #> 2 s2 right -1 s3 #> 3 s2 right -1 s3 #> 4 s3 left -1 s2 #> 5 s4 up -1 s4 #> 6 s1 down -1 s2 # Define reinforcement learning parameters control <- list(alpha = 0.1, gamma = 0.1, epsilon = 0.1) # Perform reinforcement learning model <- ReinforcementLearning(data, s = "State", a = "Action", r = "Reward", s_new = "NextState", control = control) # Print result print(model) #> State-Action function Q #> right up down left #> s1 -1.09619438 -1.098533 -1.00183072 -1.0978962 #> s2 -0.01980279 -1.097758 -1.00252228 -1.0037977 #> s3 -0.02335524 9.884394 -0.01722548 -0.9985081 #> s4 -1.09616040 -1.106392 -1.10548631 -1.1059655 #> #> Policy #> s1 s2 s3 s4 #> "down" "right" "up" "right" #> #> Reward (last iteration) #>  -263
ReinforcementLearning is released under the MIT License
Copyright (c) 2018 Nicolas Pröllochs & Stefan Feuerriegel