def nutrient_simulator(time_obj):
return nutrient_simulators.ssm_nutrient_simulator(time_obj,
out_trans_mat1=out_trans_mat1,
out_trans_mat2=out_trans_mat2,
p_switch_to_switch=p_switch_to_switch,
p_noswitch_to_switch=p_noswitch_to_switch,
p_init_output=p_init_output)
def test_a_ssm_simulator(self):
print "test ssm glucose-galactose simulator"
p_switch_to_switch = 0.10
p_noswitch_to_switch = 0.95
p_init_output = 0.5
p_init_switch = 0.5
num_iters = 5
time_obj = time_unit.Time(1, 150, 1)
out_trans_mat1 = np.array([[0., 1],
[1., 0]])
out_trans_mat2 = np.array([[1., 0.],
[1., 0.]])
for n in xrange(num_iters):
print "iter = %d" %(n)
data = \
nutrient_simulators.ssm_nutrient_simulator(time_obj,
out_trans_mat1=out_trans_mat1,
out_trans_mat2=out_trans_mat2,
p_switch_to_switch=p_switch_to_switch,
p_noswitch_to_switch=p_noswitch_to_switch,
p_init_switch=p_init_switch,
p_init_output=p_init_output)
def test_fitness_policies(self):
"""
Compare several policies to each other.
"""
###
### TODO: finish this. Make environment with mostly glu
###
# set seed
np.random.seed(20)
t_start = 0
t_end = 100
t_step = 1
time_obj = time_unit.Time(t_start, t_end, t_step)
# choose an environment that gives mostly glucose states
# and see if it leads to roughly similar population size
# that you'd get with glucose-only growth
out_trans_mat1 = np.array([[0., 1],
[1., 0]])
out_trans_mat2 = np.array([[1., 0.],
[1., 0.]])
p_switch_to_switch = 0.10
p_noswitch_to_switch = 0.10
p_init_output = 0.5
p_init_switch = 0.5
data = \
nutrient_simulators.ssm_nutrient_simulator(time_obj,
out_trans_mat1=out_trans_mat1,
out_trans_mat2=out_trans_mat2,
p_switch_to_switch=p_switch_to_switch,
p_noswitch_to_switch=p_noswitch_to_switch,
p_init_switch=p_init_switch,
p_init_output=p_init_output)
data = np.array(data)
print "data: ", data
num_timesteps = len(data)
nutr_growth_rates = [0.3, 0.075]
mismatch_growth_rate = 0.
nutr_labels = ["glucose", "galactose"]
num_nutrs = len(nutr_labels)
def nutr_simulator(time_obj):
return data
# create environment
env_obj = \
env.MixedDiscEnvironment(nutr_labels, nutr_simulator,
nutr_growth_rates=nutr_growth_rates,
mismatch_growth_rate=mismatch_growth_rate)
# simulate the environment
env_obj.simulate(time_obj)
# policies to run on environment
all_policies = OrderedDict()
all_policies["Posterior predictive"] = policies.posterior_pred_policy
# all_policies["Plastic"] = policies.plastic_growth_policy
# all_policies["Random"] = policies.rand_growth_policy
all_policies["Glucose-only"] = policies.glu_only_growth_policy
# run fitness simulation with these policies
init_pop_sizes = [100, 0]
params = {"nutr_labels": nutr_labels,
"nutr_growth_rates": nutr_growth_rates,
"mismatch_growth_rate": mismatch_growth_rate,
"decision_lag_time": 1,
"init_pop_sizes": init_pop_sizes,
"num_sim_iters": 1,
# policy-related parameters
"num_switch_states": 2,
"num_outputs": num_nutrs}
fitness_obj = fitness.FitnessSim(all_policies, env_obj, params)
fitness_results = fitness_obj.simulate(time_obj)
print "final population sizes: "
final_time = fitness_results[fitness_results["t"] == time_obj.t[-1]]
# compare the theoretical attainable population size
# based on deterministic growth model in glucose
# to the result we get: they should be similar.
post_pred_results = final_time[final_time["policy"] == "Posterior predictive"]
post_pred_popsize = post_pred_results["pop_size"]