LEARNING ENVIRONMENT FOR NB-DQN

Repository setup

ln -s /path/to/maze_generator maze_generator

Requirements

• Lua >= 5.2
• Torch7

API

-- Create new learning environment
e = Environment() -- the size of maze field is 9x9

-- Reset current environment (clear movment history)
e:reset()

-- Move agent in the maze
e:move(direction)

-- get agent's visual information about wall
e:wall() -- => [0; 0; 1; 1]

-- get the coordinate where the agent currently exists
e.current_coordinate -- => { 0, 1 }

-- get the novelty that is given for last movement
e.novelty -- => 0 or 10