def test_pick_agent_index_to_move(self):
agent_indices = get_agent_indices(self.test_array)
unsatisfied_agent_index_indices = [0, 1, 2, 3, 5, 6, 7, 8]
unsatisfied_agent_indices = \
agent_indices[unsatisfied_agent_index_indices]
expected_agent_index = _first_picker(unsatisfied_agent_indices)
agent_index = _pick_agent_index_to_move(unsatisfied_agent_indices,
_first_picker)
with self.subTest(out=agent_index, expected=expected_agent_index):
self.assertEqual(expected_agent_index, agent_index)
def check_get_unsatisfied_agent_indices_expected_output(
self, parameters, satisficers):
agent_indices = get_agent_indices(self.test_array)
for utility_function, expected_output in parameters:
utility = ut.get_utility_for_array(utility_function,
self.test_array)
output = _get_unsatisfied_agent_indices(utility,
agent_indices,
satisficers=satisficers)
expected_output = agent_indices[np.array(expected_output)]
with self.subTest(ut=utility_function,
out=output,
expected=expected_output):
self.assertTrue(np.array_equal(output, expected_output))
def test_satisfied_percent(self):
test_arrays = [(np.array([[0, 1, 1, 0], [1, 1, 0, 1], [0, 2, 2, 0],
[0, 1, 2, 0]]), 8 * 100 / 9),
(np.array([[0, 1, 2, 0], [1, 2, 2, 1], [1, 2, 2, 0],
[1, 1, 1, 0]]), 9 * 100 / 12),
(np.array([[1, 1, 1, 0], [1, 1, 0, 2], [0, 1, 2, 2],
[0, 1, 2, 2]]), 100)]
utility_function = create_flat_utility(0.5)
for array, exepcted_output in test_arrays:
ut = get_utility_for_array(utility_function, array, True)
agent_indices = get_agent_indices(array)
sp = satisfied_percent(ut, agent_indices)
with self.subTest():
self.assertEqual(sp, exepcted_output)
def check_better_vacancy_expected_output(self, parameters, satisficers):
utility = get_utility_for_array(create_flat_utility(0.5),
self.test_array)
agent_indices = get_agent_indices(self.test_array)
vacancy_indices = get_vacancy_indices(self.test_array)
for unsatisfied_agent_index, expected_output_indices in parameters:
expected_output = vacancy_indices[expected_output_indices]
i = agent_indices[unsatisfied_agent_index]
output = _get_better_vacancies(self.test_array,
i,
utility,
vacancy_indices,
satisficers=satisficers)
with self.subTest(i=unsatisfied_agent_index,
out=output,
expected=expected_output):
self.assertTrue(np.array_equal(output, expected_output))
def is_simulation_halted(array, utility_function): utility = get_utility_for_array(utility_function, array, count_vacancies=True) agent_indices = get_agent_indices(array) if agent_indices.size == 0: return True unsatisfied_agent_indices = _get_unsatisfied_agent_indices(utility, agent_indices, satisficers=False) if unsatisfied_agent_indices.size == 0: return True vacancy_indices = get_vacancy_indices(array) if vacancy_indices.size == 0: return True for agent_index in unsatisfied_agent_indices: better_vacancies = _get_better_vacancies(array, agent_index, utility, vacancy_indices, satisficers=False) if better_vacancies.size != 0: return False return True
def run_simulation(settings, callback=lambda arr, res, i: None):
"""Run simulation with specified settings.
Call the optional callback function after each iteration,
passing the current array state, current result, and iteration number.
Returns simulation result - segregation measures before each iteration.
Args:
settings (SimulationSettings): settings
callback (callable, optional): Function to call after each iteration
Returns:
SimulationResult: Result - segregation measures for each iteration
"""
global _stop
settings.validate()
result = SimulationResult(settings.segregation_measure_names)
array = create_array(settings.grid_size,
settings.get_agent_type_proportions(),
settings.initial_random_allocation)
# utility - function: (index) -> (0,1)
utility = get_utility_for_array(settings.utility_function, array,
count_vacancies=settings.count_vacancies, radius=settings.radius)
pickers = {
'random': _random_picker,
'first': _first_picker,
}
agent_pickers = dict(pickers, **{
'roulette': _create_roulette_picker(
settings.agent_roulette_base_weight, utility, for_agents=True),
})
vacancy_pickers = dict(pickers, **{
'roulette': _create_roulette_picker(
settings.vacancy_roulette_base_weight, utility, for_agents=False),
})
agent_picker = agent_pickers[settings.agent_picking_regime]
vacancy_picker = vacancy_pickers[settings.vacancy_picking_regime]
# if no agents, end simulation
agent_indices = get_agent_indices(array)
if agent_indices.shape[0] == 0:
return result
# _update_result(result, array, agent_indices,
# settings.count_vacancies, settings.segregation_measure_names)
for i in range(settings.iterations):
while _pause:
pass
should_save_result = i % settings.save_period == 0
if should_save_result:
callback(array, result, i)
is_simulation_halted = update_array(array, utility,
result, agent_picker, vacancy_picker, settings.count_vacancies,
settings.segregation_measure_names, settings.satisficers,
should_save_result)
# if no further moves are possible, exit simulation early
if is_simulation_halted:
# Get measures for final state and save (unless it was just saved)
if not should_save_result:
agent_indices = get_agent_indices(array)
_update_result(result, array, agent_indices,
settings.count_vacancies, settings.segregation_measure_names)
callback(array, result, i)
break
if _stop:
break
_stop = False
return result
def update_array(array, utility, result, agent_picker, vacancy_picker, count_vacancies, segregation_measure_names, satisficers=False, save_result=True): """Do a single iteration of the simulation. Args: array (ndarray): array utility (callable): utility function - (agent index) -> (0, 1) radius (int): neighborhood radius result (SimulationResult): result object to store segregation measures before updating array satisficers (bool, optional): satisficer behavior setting """ agent_indices = get_agent_indices(array) if save_result: _update_result(result, array, agent_indices, count_vacancies, segregation_measure_names) unsatisfied_agent_indices = _get_unsatisfied_agent_indices(utility, agent_indices, satisficers=satisficers) # if all agents satisfied, end simulation if unsatisfied_agent_indices.size == 0: return True agent_index = _pick_agent_index_to_move(unsatisfied_agent_indices, agent_picker) vacancy_indices = get_vacancy_indices(array) if vacancy_indices.size == 0: return True # if vacancy picker is roulette, consider all vacancies if vacancy_picker.__name__ == 'roulette_picker': better_vacancy_indices = vacancy_indices else: better_vacancy_indices = _get_better_vacancies(array, agent_index, utility, vacancy_indices, satisficers=satisficers) if better_vacancy_indices.size == 0: # if relocation regime is move first, and the first agent has no better # vacancy, the simulation will halt. If this happens, try all # unsatisfied agents in list and pick the first one which can find a # better vacancy. If none have a better vacancy, end the simulation. if agent_picker is _first_picker: for agent_index in unsatisfied_agent_indices: better_vacancy_indices = _get_better_vacancies(array, agent_index, utility, vacancy_indices, satisficers=satisficers) if better_vacancy_indices.size != 0: break else: return True else: return False agent_type = array[agent_index] better_vacancy = _pick_better_vacancy_index(better_vacancy_indices, vacancy_picker, agent_type) _move(array, agent_index, better_vacancy) return False
def test_agent_indices(self):
output = get_agent_indices(self.test_array)
expected_output = self.agent_indices
self.assertTrue(np.array_equal(output, expected_output))