a2rl.utils.data_generator_simple#

a2rl.utils.data_generator_simple(markov_order=0, action_effect=False, policy=False, reward_function=False)[source]#

Generate different types of data for your testing.

Parameters:

  • markov_order (int) – the order of the synthetic data

  • action_effect (bool) – allow the action to have an effect on the states

  • policy (bool) – generate the actions with some rules

  • reward_function (bool) – create a reward function with states and actions

Return type:

WiDataFrame

Reference:

if markov_order=0 then the states are randomly generated if markov_order=1 then the next state is affected by the previous one only if markov_order>1 then the next state is affected by a mixture of the previous history. Keep this number less than 10.

if action_affect = True then the actions can affect the state as well by using a different transition function

if policy = True then there is a consistent rule choosing the action

if reward_function = True then the reward function is calculated on the states