np.random.seed(99) N = 1000 # number of training points x = 100 * np.random.rand(N) f = lambda x_: 6 * np.sin(pi * x_ / 10.0) / (pi * x_ / 10.0 + 1) y_ = f(x) + np.math.sqrt(0.05)*np.random.randn(x.shape[0]) y = np.sign(y_) y[y == -1] = 0 x_test = np.linspace(np.min(x)-5.0, np.max(x)+5.0, num=500) y_test = np.sign(f(x_test) + np.math.sqrt(0.05)*np.random.randn(x_test.shape[0])) y_test[y_test == -1] = 0 var_f = 1. # GP variance len_f = 5.0 # GP lengthscale prior = priors.Matern52(variance=var_f, lengthscale=len_f) lik = likelihoods.Bernoulli(link='logit') inf_method = approx_inf.ExpectationPropagation(power=0.9, intmethod='UT') # inf_method = approx_inf.VariationalInference(intmethod='GH') # inf_method = approx_inf.VariationalInference(intmethod='UT') # inf_method = approx_inf.ExtendedEP(power=0) # inf_method = approx_inf.ExtendedKalmanSmoother() # inf_method = approx_inf.GaussHermiteKalmanSmoother() # inf_method = approx_inf.StatisticallyLinearisedEP(intmethod='UT') # inf_method = approx_inf.UnscentedKalmanSmoother() model = SDEGP(prior=prior, likelihood=lik, t=x, y=y, approx_inf=inf_method) opt_init, opt_update, get_params = optimizers.adam(step_size=2e-1) # parameters should be a 2-element list [param_prior, param_likelihood]
print('method number', method)
print('batch number', fold)
# Get training and test indices
ind_test = ind_split[fold] # np.sort(ind_shuffled[:N//10])
ind_train = np.concatenate(ind_split[np.arange(10) != fold])
x_train = x # [ind_train] # 90/10 train/test split
x_test = x # [ind_test]
y_train = y # [ind_train]
y_test = y # [ind_test]
N_batch = 5000
M = 5000
# z = np.linspace(701050, 737050, M)
z = np.linspace(x[0], x[-1], M)
prior_1 = priors.Matern52(variance=2., lengthscale=5.5e4)
prior_2 = priors.QuasiPeriodicMatern32(variance=1., lengthscale_periodic=2., period=365., lengthscale_matern=1.5e4)
prior_3 = priors.QuasiPeriodicMatern32(variance=1., lengthscale_periodic=2., period=7., lengthscale_matern=30*365.)
prior = priors.Sum([prior_1, prior_2, prior_3])
lik = likelihoods.Poisson()
if method == 0:
inf_method = approx_inf.EKS(damping=.5)
elif method == 1:
inf_method = approx_inf.UKS(damping=.5)
elif method == 2:
inf_method = approx_inf.GHKS(damping=.5)
elif method == 3:
inf_method = approx_inf.EP(power=1, intmethod='GH', damping=.5)
elif method == 4:
N = 1000
x = np.sort(
np.random.permutation(
np.linspace(-25.0, 150.0, num=N) +
0.5 * np.random.randn(N))) # unevenly spaced
x_test = np.linspace(np.min(x) - 15.0, np.max(x) + 15.0, num=100)
dummy_y = x
var_f = 1.0 # GP variance
len_f = 20.0 # GP lengthscale
var_y = 0.1
theta_prior = jnp.array([var_f, len_f])
theta_lik = jnp.array([])
prior_ = priors.Matern52(theta_prior)
lik_ = likelihoods.SumOfGaussians(theta_lik)
approx_inf_ = EP(power=1.)
# approx_inf_ = PL()
# approx_inf_ = CL(power=0.5)
# approx_inf_ = IKS()
# approx_inf_ = EKEP()
sde_gp_model = SDEGP(prior=prior_,
likelihood=lik_,
t=x,
y=dummy_y,
t_test=x_test,
approx_inf=approx_inf_)
print('generating some data by sampling from the prior ...')