def test_simple_svd(self):
pm = SVDMapper()
# train SVD
pm.train(self.ndlin)
self.failUnlessEqual(pm.proj.shape, (20, 20))
# now project data into PCA space
p = pm.forward(self.ndlin)
# only first eigenvalue significant
self.failUnless(pm.sv[:1] > 1.0)
self.failUnless((pm.sv[1:] < 0.0001).all())
# only variance of first component significant
var = p.var(axis=0)
# test that only one component has variance
self.failUnless(var[:1] > 1.0)
self.failUnless((var[1:] < 0.0001).all())
# check that the mapped data can be fully recovered by 'reverse()'
pr = pm.reverse(p)
self.failUnlessEqual(pr.shape, (40, 20))
self.failUnless(np.abs(pm.reverse(p) - self.ndlin).sum() < 0.0001)
def test_simple_svd(self):
pm = SVDMapper()
# train SVD
pm.train(self.ndlin)
self.failUnlessEqual(pm.proj.shape, (20, 20))
# now project data into PCA space
p = pm.forward(self.ndlin)
# only first eigenvalue significant
self.failUnless(pm.sv[:1] > 1.0)
self.failUnless((pm.sv[1:] < 0.0001).all())
# only variance of first component significant
var = p.var(axis=0)
# test that only one component has variance
self.failUnless(var[:1] > 1.0)
self.failUnless((var[1:] < 0.0001).all())
# check that the mapped data can be fully recovered by 'reverse()'
pr = pm.reverse(p)
self.failUnlessEqual(pr.shape, (40,20))
self.failUnless(np.abs(pm.reverse(p) - self.ndlin).sum() < 0.0001)
def test_more_svd(self):
pm = SVDMapper()
# train SVD
pm.train(self.largefeat)
# mixing matrix cannot be square
self.failUnlessEqual(pm.proj.shape, (40, 10))
# only first singular value significant
self.failUnless(pm.sv[:1] > 10)
self.failUnless((pm.sv[1:] < 10).all())
# now project data into SVD space
p = pm.forward(self.largefeat)
# only variance of first component significant
var = p.var(axis=0)
# test that only one component has variance
self.failUnless(var[:1] > 1.0)
self.failUnless((var[1:] < 0.0001).all())
# check that the mapped data can be fully recovered by 'reverse()'
rp = pm.reverse(p)
self.failUnlessEqual(rp.shape, self.largefeat.shape)
self.failUnless((np.round(rp) == self.largefeat).all())
# copy mapper
pm2 = deepcopy(pm)
# now make new random data and do forward->reverse check
data = np.random.normal(size=(98, 40))
data_f = pm.forward(data)
self.failUnlessEqual(data_f.shape, (98, 10))
data_r = pm.reverse(data_f)
self.failUnlessEqual(data_r.shape, (98, 40))
def test_more_svd(self):
pm = SVDMapper()
# train SVD
pm.train(self.largefeat)
# mixing matrix cannot be square
self.failUnlessEqual(pm.proj.shape, (40, 10))
# only first singular value significant
self.failUnless(pm.sv[:1] > 10)
self.failUnless((pm.sv[1:] < 10).all())
# now project data into SVD space
p = pm.forward(self.largefeat)
# only variance of first component significant
var = p.var(axis=0)
# test that only one component has variance
self.failUnless(var[:1] > 1.0)
self.failUnless((var[1:] < 0.0001).all())
# check that the mapped data can be fully recovered by 'reverse()'
rp = pm.reverse(p)
self.failUnlessEqual(rp.shape, self.largefeat.shape)
self.failUnless((np.round(rp) == self.largefeat).all())
# copy mapper
pm2 = deepcopy(pm)
# now make new random data and do forward->reverse check
data = np.random.normal(size=(98,40))
data_f = pm.forward(data)
self.failUnlessEqual(data_f.shape, (98,10))
data_r = pm.reverse(data_f)
self.failUnlessEqual(data_r.shape, (98,40))
from mvpa.mappers.mdp_adaptor import ICAMapper, PCAMapper
from mvpa import cfg
center = [10, 20]
axis_range = 7
##REF: Name was automagically refactored
def plot_proj_dir(p):
pl.plot([0, p[0, 0]], [0, p[0, 1]], linewidth=3, hold=True, color='y')
pl.plot([0, p[1, 0]], [0, p[1, 1]], linewidth=3, hold=True, color='k')
mappers = {
'PCA': PCAMapper(),
'SVD': SVDMapper(),
'ICA': ICAMapper(alg='CuBICA'),
}
datasets = [
noisy_2d_fx(100, lambda x: x, [lambda x: x], center, noise_std=0.5),
noisy_2d_fx(50,
lambda x: x, [lambda x: x, lambda x: -x],
center,
noise_std=0.5),
noisy_2d_fx(50,
lambda x: x, [lambda x: x, lambda x: 0],
center,
noise_std=0.5),
]
ndatasets = len(datasets)
def testMoreSVD(self):
pm = SVDMapper()
# train SVD
pm.train(self.largefeat)
# mixing matrix cannot be square
self.failUnlessEqual(pm.proj.shape, (40, 10))
# only first singular value significant
self.failUnless(pm.sv[:1] > 10)
self.failUnless((pm.sv[1:] < 10).all())
# now project data into SVD space
p = pm.forward(self.largefeat.samples)
# only variance of first component significant
var = p.var(axis=0)
# test that only one component has variance
self.failUnless(var[:1] > 1.0)
self.failUnless((var[1:] < 0.0001).all())
# check that the mapped data can be fully recovered by 'reverse()'
rp = pm.reverse(p)
self.failUnlessEqual(rp.shape, self.largefeat.samples.shape)
self.failUnless((N.round(rp) == self.largefeat.samples).all())
self.failUnlessEqual(pm.getInSize(), 40)
self.failUnlessEqual(pm.getOutSize(), 10)
# copy mapper
pm2 = deepcopy(pm)
# now remove all but the first 2 components from the mapper
pm2.selectOut([0, 1])
# sanity check
self.failUnlessEqual(pm2.getInSize(), 40)
self.failUnlessEqual(pm2.getOutSize(), 2)
# but orginal mapper must be left intact
self.failUnlessEqual(pm.getInSize(), 40)
self.failUnlessEqual(pm.getOutSize(), 10)
# data should still be fully recoverable by 'reverse()'
rp2 = pm2.reverse(p[:, [0, 1]])
self.failUnlessEqual(rp2.shape, self.largefeat.samples.shape)
self.failUnless(N.abs(rp2 - self.largefeat.samples).sum() < 0.0001)
# now make new random data and do forward->reverse check
data = N.random.normal(size=(98, 40))
data_f = pm.forward(data)
self.failUnlessEqual(data_f.shape, (98, 10))
data_r = pm.reverse(data_f)
self.failUnlessEqual(data_r.shape, (98, 40))
