Simple program to practice reinforcement learning
A simple game designed to understand reinforcement learning using Q-Table
reach_circle_commandline.py -
Extremely fast exploration with fixed starting point but no visuals
reach_circle_commandline_test.py -
Visualize the post reinforcement learning performance from above script
reach_circle_fromanywhere_cl.py -
Same as reach_circle_commandline.py but multiple possible starting points
reach_circle_fromanywhere_test.py -
Visualize the post reinforcement learning performance from multi startpoint script above
reach_circle_explore_exploit.py -
Visualize exploration followed by gradually increasing exploitation
strc_commandline.py -
This the reach_circle_commandline script extended to 2 balls. Green ball has to evade red ball and reach goal
strc_commandline_test.py -
This the reach_circle_commandline_test script extended to 2 balls. Green ball has to evade red ball and reach goal
Key parameters:
alpha - learning rate
epsilon - exploration threshold
gamma - discount factor for rewards from events far in time