def test_SIR_abc_SquareLoss(self):
y = self.solution[1::, 1:3]
# setting the parameters in the inference
parameters = [
pgabc.Parameter('beta', 'unif', 0, 3, logscale=False),
pgabc.Parameter('gamma', 'unif', 0, 3, logscale=False)
]
# creating the loss and abc objects
sir_obj = pgabc.create_loss("SquareLoss", parameters, self.ode,
self.x0, self.t[0], self.t[1::], y,
['I', 'R'])
sir_abc = pgabc.ABC(sir_obj, parameters)
# getting the posterior sample
sir_abc.get_posterior_sample(N=100, tol=np.inf, G=10, q=0.5)
sir_abc.continue_posterior_sample(N=100,
tol=sir_abc.next_tol,
G=10,
q=0.5)
# the estimate for beta must be between 0.485 and 0.515
# the estimate for gamma must be between 0.32 and 0.3466
med_est = np.median(sir_abc.res, axis=0)
self.assertTrue(np.allclose(med_est, self.target, 1e-2, 1e-2))
def test_SIR_abc_NormalLoss(self):
y = self.solution[1::, 1:3]
parameters = [pgabc.Parameter('beta', 'unif', 0, 3, logscale=False),
pgabc.Parameter('gamma', 'unif', 0, 3, logscale=False)]
sir_obj = pgabc.create_loss(NormalLoss, parameters, self.ode, self.x0, self.t[0],
self.t[1::], y, ['I', 'R'])
sir_abc = pgabc.ABC(sir_obj, parameters)
sir_abc.get_posterior_sample(N=100, tol=np.inf, G=10, q=0.5)
sir_abc.continue_posterior_sample(N=100, tol=sir_abc.next_tol, G=10, q=0.5)
med_est = np.median(sir_abc.res, axis=0)
self.assertTrue(np.allclose(med_est, self.target, 1e-2, 1e-2))