def init_simulation_huge_rewards():
model_object = get_model_obj(
get_random_model_specification(
constr={
"num_types":
4,
"type_spec_shifts":
np.zeros(12).reshape(4, 3),
"coeffs_work":
np.concatenate((np.random.uniform(size=10),
np.array([2000000000000000, 0, 0]))),
}))
dat = simulate(model_object)
agent_to_check = np.random.randint(0, model_object.num_agents_sim)
period_to_check = np.random.randint(0, model_object.num_periods - 1)
# print(dat[num_periods*(agent_to_check-1)+period_to_check,2])
return dat, model_object.num_periods, period_to_check, agent_to_check
def init_simulation(constr=False):
model_object = get_model_obj(
get_random_model_specification(
constr={
"num_types": 4,
"type_spec_shifts": np.zeros(12).reshape(4, 3)
}))
dat = simulate(model_object)
agent_to_check = np.random.randint(0, model_object.num_agents_sim)
period_to_check = np.random.randint(0, model_object.num_periods - 1)
# print(dat[num_periods*(agent_to_check-1)+period_to_check,2])
return (
dat,
model_object.num_periods,
model_object.num_agents_sim,
period_to_check,
model_object.delta,
agent_to_check,
)
def set_up_last_period():
# We want to set up a basic testing infrastructure for the state space creation.
model_object = get_model_obj(
get_random_model_specification(
constr={
"num_types": 4,
"type_spec_shifts": np.zeros(12).reshape(4, 3)
}))
state_space = create_state_space(model_object)
immediate_rewards = return_immediate_rewards(model_object, state_space)
periods_draws_emax = return_simulated_shocks(model_object,
simulation=False)
periods_emax = backward_induction_procedure(model_object, state_space,
immediate_rewards,
periods_draws_emax)
k_to_check = np.random.randint(1, state_space["states_number_period"][-1])
draws_to_check = periods_draws_emax[model_object.num_periods - 1]
immediate_rewards_last_period = immediate_rewards[
model_object.num_periods - 1, k_to_check]
state_to_check = state_space["states_all"][model_object.num_periods - 1,
k_to_check]
aux = np.zeros(model_object.num_draws_emax * 3).reshape(
model_object.num_draws_emax, 3)
# des is der common anteil
common = 0
if state_to_check[1] >= 12:
common = model_object.coeffs_common[0] + common
else:
common = common
if state_to_check[1] >= 15:
common = model_object.coeffs_common[1] + common
else:
common = common
# des is der genral teil
general = model_object.coeffs_work[10]
if state_to_check[0] == 0:
general = general + model_object.coeffs_work[12]
elif state_to_check[2] != 1:
general = general + model_object.coeffs_work[11]
else:
general = general
for x in range(model_object.num_draws_emax):
aux[x, 0] = ((immediate_rewards_last_period[0] - common - general) *
draws_to_check[x, 0] + common + general)
aux[x, 1] = immediate_rewards_last_period[1] + draws_to_check[x, 1]
aux[x, 2] = immediate_rewards_last_period[2] + draws_to_check[x, 2]
out = np.zeros(model_object.num_draws_emax)
for i in range(model_object.num_draws_emax):
out[i] = aux[i, :].max()
manual_result = out.sum() / model_object.num_draws_emax
return manual_result, periods_emax, model_object.num_periods, k_to_check
def random_model_object():
model_object = get_model_obj(get_random_model_specification())
return model_object
def set_up_any_period():
model_object = get_model_obj(
get_random_model_specification(
constr={
"num_types": 4,
"type_spec_shifts": np.zeros(12).reshape(4, 3)
}))
state_space = create_state_space(model_object)
immediate_rewards = return_immediate_rewards(model_object, state_space)
periods_draws_emax = return_simulated_shocks(model_object,
simulation=False)
periods_emax = backward_induction_procedure(model_object, state_space,
immediate_rewards,
periods_draws_emax)
period_to_check = np.random.randint(0, model_object.num_periods - 1)
k_to_check = np.random.randint(
0, state_space["states_number_period"][period_to_check] - 1)
draws_to_check = periods_draws_emax[period_to_check]
immediate_rewards_last_period = immediate_rewards[period_to_check,
k_to_check]
state_to_check = state_space["states_all"][period_to_check, k_to_check]
aux = np.zeros(model_object.num_draws_emax * 3).reshape(
model_object.num_draws_emax, 3)
# des is der common anteil
common = 0
if state_to_check[1] >= 12:
common = model_object.coeffs_common[0] + common
else:
common = common
if state_to_check[1] >= 15:
common = model_object.coeffs_common[1] + common
else:
common = common
# des is der genral teil
general = model_object.coeffs_work[10]
if state_to_check[0] == 0:
general = general + model_object.coeffs_work[12]
elif state_to_check[2] != 1:
general = general + model_object.coeffs_work[11]
else:
general = general
##Obtain next periods states
next_period_1 = state_space["mapping_state_idx"][period_to_check + 1,
state_to_check[0] + 1,
state_to_check[1], 0,
state_to_check[3] - 1, ]
next_period_2 = state_space["mapping_state_idx"][period_to_check + 1,
state_to_check[0],
state_to_check[1] + 1, 1,
state_to_check[3] - 1, ]
next_period_3 = state_space["mapping_state_idx"][period_to_check + 1,
state_to_check[0],
state_to_check[1], 2,
state_to_check[3] - 1, ]
for x in range(model_object.num_draws_emax):
aux[x, 0] = ((immediate_rewards_last_period[0] - common - general) *
draws_to_check[x, 0] + common + general +
model_object.delta *
periods_emax[period_to_check + 1, next_period_1])
aux[x, 1] = (immediate_rewards_last_period[1] + draws_to_check[x, 1] +
model_object.delta *
periods_emax[period_to_check + 1, next_period_2])
aux[x, 2] = (immediate_rewards_last_period[2] + draws_to_check[x, 2] +
model_object.delta *
periods_emax[period_to_check + 1, next_period_3])
out = np.zeros(model_object.num_draws_emax)
for i in range(model_object.num_draws_emax):
out[i] = aux[i, :].max()
manual_result = out.sum() / model_object.num_draws_emax
return manual_result, periods_emax, period_to_check, k_to_check