apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-15 06:52:39 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

merging with the gpy devel branch to be in sync with the latest code and make pull request again ..

Browse source
This commit is contained in:
Akash Kumar Dhaka 2017-08-10 11:47:26 +03:00
commit f0f1a183b0
12 changed files with 204 additions and 40 deletions
Download patch file Download diff file Expand all files Collapse all files

2
GPy/__version__.py
View file

@ -1 +1 @@
__version__ = "1.6.2"
__version__ = "1.7.7"

3
GPy/likelihoods/binomial.py
View file

@ -124,7 +124,7 @@ class Binomial(Likelihood):
"""
N = Y_metadata['trials']
np.testing.assert_array_equal(N.shape, y.shape)
Ny = N-y
t1 = np.zeros(y.shape)
t2 = np.zeros(y.shape)
@ -132,6 +132,7 @@ class Binomial(Likelihood):
t2[Ny>0] = -(Ny[Ny>0])/np.square(1.-inv_link_f[Ny>0])
return t1+t2
def d3logpdf_dlink3(self, inv_link_f, y, Y_metadata=None):
"""
Third order derivative log-likelihood function at y given inverse link of f w.r.t inverse link of f

101
GPy/testing/gpy_kernels_state_space_tests.py
View file

@ -306,11 +306,7 @@ class StateSpaceKernelsTests(np.testing.TestCase):
gp_kernel=gp_kernel,
mean_compare_decimal=2, var_compare_decimal=2)
def test_forecast(self,):
"""
Test time-series forecasting.
"""
def test_forecast_regular(self,):
# Generate data ->
np.random.seed(339) # seed the random number generator
#import pdb; pdb.set_trace()
@ -334,37 +330,102 @@ class StateSpaceKernelsTests(np.testing.TestCase):
#import pdb; pdb.set_trace()
def get_new_kernels():
periodic_kernel = GPy.kern.StdPeriodic(1,active_dims=[0,])
gp_kernel = GPy.kern.Linear(1, active_dims=[0,]) + GPy.kern.Bias(1, active_dims=[0,]) + periodic_kernel
gp_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
gp_kernel.std_periodic.period.constrain_bounded(0.15, 100)
periodic_kernel = GPy.kern.StdPeriodic(1,active_dims=[0,])
gp_kernel = GPy.kern.Linear(1, active_dims=[0,]) + GPy.kern.Bias(1, active_dims=[0,]) + periodic_kernel
gp_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
gp_kernel.std_periodic.period.constrain_bounded(0.15, 100)
periodic_kernel = GPy.kern.sde_StdPeriodic(1,active_dims=[0,])
ss_kernel = GPy.kern.sde_Linear(1,X,active_dims=[0,]) + \
GPy.kern.sde_Bias(1, active_dims=[0,]) + periodic_kernel
periodic_kernel = GPy.kern.sde_StdPeriodic(1,active_dims=[0,])
ss_kernel = GPy.kern.sde_Linear(1,X,active_dims=[0,]) + \
GPy.kern.sde_Bias(1, active_dims=[0,]) + periodic_kernel
ss_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
ss_kernel.std_periodic.period.constrain_bounded(0.15, 100)
ss_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
ss_kernel.std_periodic.period.constrain_bounded(0.15, 100)
return ss_kernel, gp_kernel
ss_kernel, gp_kernel = get_new_kernels()
self.run_for_model(X_train, Y_train, ss_kernel, kalman_filter_type = 'regular',
use_cython=False, optimize_max_iters=30, check_gradients=True,
predict_X=X_test,
gp_kernel=gp_kernel,
mean_compare_decimal=2, var_compare_decimal=2)
def test_forecast_svd(self,):
# Generate data ->
np.random.seed(339) # seed the random number generator
#import pdb; pdb.set_trace()
(X,Y) = generate_sine_data(x_points=None, sin_period=5.0, sin_ampl=5.0, noise_var=2.0,
plot = False, points_num=100, x_interval = (0, 40), random=True)
(X1,Y1) = generate_linear_data(x_points=X, tangent=1.0, add_term=20.0, noise_var=0.0,
plot = False, points_num=100, x_interval = (0, 40), random=True)
Y = Y + Y1
X_train = X[X <= 20]
Y_train = Y[X <= 20]
X_test = X[X > 20]
Y_test = Y[X > 20]
X.shape = (X.shape[0],1); Y.shape = (Y.shape[0],1)
X_train.shape = (X_train.shape[0],1); Y_train.shape = (Y_train.shape[0],1)
X_test.shape = (X_test.shape[0],1); Y_test.shape = (Y_test.shape[0],1)
# Generate data <-
#import pdb; pdb.set_trace()
periodic_kernel = GPy.kern.StdPeriodic(1,active_dims=[0,])
gp_kernel = GPy.kern.Linear(1, active_dims=[0,]) + GPy.kern.Bias(1, active_dims=[0,]) + periodic_kernel
gp_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
gp_kernel.std_periodic.period.constrain_bounded(0.15, 100)
periodic_kernel = GPy.kern.sde_StdPeriodic(1,active_dims=[0,])
ss_kernel = GPy.kern.sde_Linear(1,X,active_dims=[0,]) + \
GPy.kern.sde_Bias(1, active_dims=[0,]) + periodic_kernel
ss_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
ss_kernel.std_periodic.period.constrain_bounded(0.15, 100)
ss_kernel, gp_kernel = get_new_kernels()
self.run_for_model(X_train, Y_train, ss_kernel, kalman_filter_type = 'svd',
use_cython=False, optimize_max_iters=30, check_gradients=False,
predict_X=X_test,
gp_kernel=gp_kernel,
mean_compare_decimal=2, var_compare_decimal=2)
ss_kernel, gp_kernel = get_new_kernels()
def test_forecast_svd_cython(self,):
# Generate data ->
np.random.seed(339) # seed the random number generator
#import pdb; pdb.set_trace()
(X,Y) = generate_sine_data(x_points=None, sin_period=5.0, sin_ampl=5.0, noise_var=2.0,
plot = False, points_num=100, x_interval = (0, 40), random=True)
(X1,Y1) = generate_linear_data(x_points=X, tangent=1.0, add_term=20.0, noise_var=0.0,
plot = False, points_num=100, x_interval = (0, 40), random=True)
Y = Y + Y1
X_train = X[X <= 20]
Y_train = Y[X <= 20]
X_test = X[X > 20]
Y_test = Y[X > 20]
X.shape = (X.shape[0],1); Y.shape = (Y.shape[0],1)
X_train.shape = (X_train.shape[0],1); Y_train.shape = (Y_train.shape[0],1)
X_test.shape = (X_test.shape[0],1); Y_test.shape = (Y_test.shape[0],1)
# Generate data <-
#import pdb; pdb.set_trace()
periodic_kernel = GPy.kern.StdPeriodic(1,active_dims=[0,])
gp_kernel = GPy.kern.Linear(1, active_dims=[0,]) + GPy.kern.Bias(1, active_dims=[0,]) + periodic_kernel
gp_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
gp_kernel.std_periodic.period.constrain_bounded(0.15, 100)
periodic_kernel = GPy.kern.sde_StdPeriodic(1,active_dims=[0,])
ss_kernel = GPy.kern.sde_Linear(1,X,active_dims=[0,]) + \
GPy.kern.sde_Bias(1, active_dims=[0,]) + periodic_kernel
ss_kernel.std_periodic.lengthscale.constrain_bounded(0.25, 1000)
ss_kernel.std_periodic.period.constrain_bounded(0.15, 100)
self.run_for_model(X_train, Y_train, ss_kernel, kalman_filter_type = 'svd',
use_cython=True, optimize_max_iters=30, check_gradients=False,
predict_X=X_test,

1
GPy/testing/inference_tests.py
View file

@ -85,6 +85,7 @@ class InferenceGPEP(unittest.TestCase):
inference_method=inf,
likelihood=lik)
K = self.model.kern.K(X)
post_params, ga_approx, cav_params, log_Z_tilde = self.model.inference_method.expectation_propagation(K, ObsAr(Y), lik, None)
mu_tilde = ga_approx.v / ga_approx.tau.astype(float)

9
GPy/util/datasets.py
View file

@ -206,7 +206,10 @@ def authorize_download(dataset_name=None):
def download_data(dataset_name=None):
"""Check with the user that the are happy with terms and conditions for the data set, then download it."""
import itertools
try:
from itertools import zip_longest
except ImportError:
from itertools import izip_longest as zip_longest
dr = data_resources[dataset_name]
if not authorize_download(dataset_name):
@ -220,8 +223,8 @@ def download_data(dataset_name=None):
if 'suffices' in dr: zip_urls += (dr['suffices'], )
else: zip_urls += ([],)
for url, files, save_names, suffices in itertools.zip_longest(*zip_urls, fillvalue=[]):
for f, save_name, suffix in itertools.zip_longest(files, save_names, suffices, fillvalue=None):
for url, files, save_names, suffices in zip_longest(*zip_urls, fillvalue=[]):
for f, save_name, suffix in zip_longest(files, save_names, suffices, fillvalue=None):
download_url(os.path.join(url,f), dataset_name, save_name, suffix=suffix)
return True