apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-18 13:55:14 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

expectation test slicing

Browse source
This commit is contained in:
Max Zwiessele 2013-09-03 10:06:14 +01:00
parent 1d37b31bea
commit b42a6da835
1 changed files with 36 additions and 19 deletions
Download patch file Download diff file Expand all files Collapse all files

55
GPy/testing/psi_stat_expectation_tests.py
View file

@ -7,9 +7,14 @@ import unittest
import GPy import GPy
import numpy as np import numpy as np
from GPy import testing from GPy import testing
import sys
import numpy
from GPy.kern.parts.rbf import RBF
from GPy.kern.parts.linear import Linear
from copy import deepcopy
__test__ = False __test__ = lambda: 'deep' in sys.argv
np.random.seed(0) # np.random.seed(0)
def ard(p): def ard(p):
try: try:
@ -19,24 +24,37 @@ def ard(p):
pass pass
return "" return ""
@testing.deepTest(__test__) @testing.deepTest(__test__())
class Test(unittest.TestCase): class Test(unittest.TestCase):
input_dim = 9 input_dim = 9
num_inducing = 4 num_inducing = 4
N = 3 N = 3
Nsamples = 6e6 Nsamples = 5e6
def setUp(self): def setUp(self):
i_s_dim_list = [2,4,3]
indices = numpy.cumsum(i_s_dim_list).tolist()
input_slices = [slice(a,b) for a,b in zip([None]+indices, indices)]
#input_slices[2] = deepcopy(input_slices[1])
input_slice_kern = GPy.kern.kern(9,
[
RBF(i_s_dim_list[0], np.random.rand(), np.random.rand(i_s_dim_list[0]), ARD=True),
RBF(i_s_dim_list[1], np.random.rand(), np.random.rand(i_s_dim_list[1]), ARD=True),
Linear(i_s_dim_list[2], np.random.rand(i_s_dim_list[2]), ARD=True)
],
input_slices = input_slices
)
self.kerns = ( self.kerns = (
input_slice_kern,
# (GPy.kern.rbf(self.input_dim, ARD=True) + # (GPy.kern.rbf(self.input_dim, ARD=True) +
# GPy.kern.linear(self.input_dim, ARD=True) + # GPy.kern.linear(self.input_dim, ARD=True) +
# GPy.kern.bias(self.input_dim) + # GPy.kern.bias(self.input_dim) +
# GPy.kern.white(self.input_dim)), # GPy.kern.white(self.input_dim)),
(GPy.kern.rbf(self.input_dim, np.random.rand(), np.random.rand(self.input_dim), ARD=True) + # (GPy.kern.rbf(self.input_dim, np.random.rand(), np.random.rand(self.input_dim), ARD=True) +
GPy.kern.rbf(self.input_dim, np.random.rand(), np.random.rand(self.input_dim), ARD=True) + # GPy.kern.rbf(self.input_dim, np.random.rand(), np.random.rand(self.input_dim), ARD=True) +
GPy.kern.linear(self.input_dim, np.random.rand(self.input_dim), ARD=True) + # GPy.kern.linear(self.input_dim, np.random.rand(self.input_dim), ARD=True) +
GPy.kern.bias(self.input_dim) + # GPy.kern.bias(self.input_dim) +
GPy.kern.white(self.input_dim)), # GPy.kern.white(self.input_dim)),
# GPy.kern.rbf(self.input_dim), GPy.kern.rbf(self.input_dim, ARD=True), # GPy.kern.rbf(self.input_dim), GPy.kern.rbf(self.input_dim, ARD=True),
# GPy.kern.linear(self.input_dim, ARD=False), GPy.kern.linear(self.input_dim, ARD=True), # GPy.kern.linear(self.input_dim, ARD=False), GPy.kern.linear(self.input_dim, ARD=True),
# GPy.kern.linear(self.input_dim) + GPy.kern.bias(self.input_dim), # GPy.kern.linear(self.input_dim) + GPy.kern.bias(self.input_dim),
@ -61,22 +79,22 @@ class Test(unittest.TestCase):
def test_psi1(self): def test_psi1(self):
for kern in self.kerns: for kern in self.kerns:
Nsamples = 100 Nsamples = np.floor(self.Nsamples/300.)
psi1 = kern.psi1(self.Z, self.q_x_mean, self.q_x_variance) psi1 = kern.psi1(self.Z, self.q_x_mean, self.q_x_variance)
K_ = np.zeros((Nsamples, self.num_inducing)) K_ = np.zeros((Nsamples, self.num_inducing))
diffs = [] diffs = []
for i, q_x_sample_stripe in enumerate(np.array_split(self.q_x_samples, self.Nsamples / Nsamples)): for i, q_x_sample_stripe in enumerate(np.array_split(self.q_x_samples, self.Nsamples / Nsamples)):
K = kern.K(q_x_sample_stripe, self.Z) K = kern.K(q_x_sample_stripe[:Nsamples], self.Z)
K_ += K K_ += K
diffs.append(((psi1 - (K_ / (i + 1)))).mean()) diffs.append((np.abs(psi1 - (K_ / (i + 1)))**2).mean())
K_ /= self.Nsamples / Nsamples K_ /= self.Nsamples / Nsamples
msg = "psi1: " + "+".join([p.name + ard(p) for p in kern.parts]) msg = "psi1: " + "+".join([p.name + ard(p) for p in kern.parts])
try: try:
import pylab import pylab
pylab.figure(msg) pylab.figure(msg)
pylab.plot(diffs) pylab.plot(diffs)
self.assertTrue(np.allclose(psi1.squeeze(), K_, # print msg, ((psi1.squeeze() - K_)**2).mean() < .01
rtol=1e-1, atol=.1), self.assertTrue(((psi1.squeeze() - K_)**2).mean() < .01,
msg=msg + ": not matching") msg=msg + ": not matching")
# sys.stdout.write(".") # sys.stdout.write(".")
except: except:
@ -87,7 +105,7 @@ class Test(unittest.TestCase):
def test_psi2(self): def test_psi2(self):
for kern in self.kerns: for kern in self.kerns:
Nsamples = 100 Nsamples = self.Nsamples/300.
psi2 = kern.psi2(self.Z, self.q_x_mean, self.q_x_variance) psi2 = kern.psi2(self.Z, self.q_x_mean, self.q_x_variance)
K_ = np.zeros((self.num_inducing, self.num_inducing)) K_ = np.zeros((self.num_inducing, self.num_inducing))
diffs = [] diffs = []
@ -95,13 +113,14 @@ class Test(unittest.TestCase):
K = kern.K(q_x_sample_stripe, self.Z) K = kern.K(q_x_sample_stripe, self.Z)
K = (K[:, :, None] * K[:, None, :]).mean(0) K = (K[:, :, None] * K[:, None, :]).mean(0)
K_ += K K_ += K
diffs.append(((psi2 - (K_ / (i + 1)))).mean()) diffs.append(((psi2 - (K_ / (i + 1)))**2).mean())
K_ /= self.Nsamples / Nsamples K_ /= self.Nsamples / Nsamples
msg = "psi2: {}".format("+".join([p.name + ard(p) for p in kern.parts])) msg = "psi2: {}".format("+".join([p.name + ard(p) for p in kern.parts]))
try: try:
import pylab import pylab
pylab.figure(msg) pylab.figure(msg)
pylab.plot(diffs) pylab.plot(diffs)
# print msg, np.allclose(psi2.squeeze(), K_, rtol=1e-1, atol=.1)
self.assertTrue(np.allclose(psi2.squeeze(), K_, self.assertTrue(np.allclose(psi2.squeeze(), K_,
rtol=1e-1, atol=.1), rtol=1e-1, atol=.1),
msg=msg + ": not matching") msg=msg + ": not matching")
@ -114,10 +133,8 @@ class Test(unittest.TestCase):
pass pass
if __name__ == "__main__": if __name__ == "__main__":
import sys
__test__ = 'deep' in sys.argv
sys.argv = ['', sys.argv = ['',
'Test.test_psi0', #'Test.test_psi0',
'Test.test_psi1', 'Test.test_psi1',
'Test.test_psi2', 'Test.test_psi2',
] ]