def test_nested_select(self):
def beales(x, y):
e1 = 1.5 - x + x * y
e2 = 2.25 - x + x * (y**2)
e3 = 2.625 - x + x * (y**3)
return {'e1': e1, 'e2': e2, 'e3': e3}
x1 = ch.zeros(10)
y1 = ch.zeros(10)
# With a single select this worked
minimize(beales(x1, y1),
x0=[x1[1:4], y1],
method='dogleg',
options={'disp': False})
x2 = ch.zeros(10)
y2 = ch.zeros(10)
# But this used to raise `AttributeError: 'Select' object has no attribute 'x'`
minimize(beales(x2, y2),
x0=[x2[1:8][:3], y2],
method='dogleg',
options={'disp': False})
np.testing.assert_array_equal(x1, x2)
np.testing.assert_array_equal(y1, y2)
def test_nested_select(self):
def beales(x, y):
e1 = 1.5 - x + x*y
e2 = 2.25 - x + x*(y**2)
e3 = 2.625 - x + x*(y**3)
return {'e1': e1, 'e2': e2, 'e3': e3}
x1 = ch.zeros(10)
y1 = ch.zeros(10)
# With a single select this worked
minimize(beales(x1, y1), x0=[x1[1:4], y1], method='dogleg', options={'disp': False})
x2 = ch.zeros(10)
y2 = ch.zeros(10)
# But this used to raise `AttributeError: 'Select' object has no attribute 'x'`
minimize(beales(x2, y2), x0=[x2[1:8][:3], y2], method='dogleg', options={'disp': False})
np.testing.assert_array_equal(x1, x2)
np.testing.assert_array_equal(y1, y2)
def moment(a, moment=1, axis=0):
if moment == 1:
# By definition the first moment about the mean is 0.
shape = list(a.shape)
del shape[axis]
if shape:
# return an actual array of the appropriate shape
return ch.zeros(shape, dtype=float)
else:
# the input was 1D, so return a scalar instead of a rank-0 array
return np.float64(0.0)
else:
mn = ch.expand_dims(a.mean(axis=axis), axis)
s = ch.power((a - mn), moment)
return s.mean(axis=axis)
def moment(a, moment=1, axis=0):
if moment == 1:
# By definition the first moment about the mean is 0.
shape = list(a.shape)
del shape[axis]
if shape:
# return an actual array of the appropriate shape
return ch.zeros(shape, dtype=float)
else:
# the input was 1D, so return a scalar instead of a rank-0 array
return np.float64(0.0)
else:
mn = ch.expand_dims(a.mean(axis=axis), axis)
s = ch.power((a-mn), moment)
return s.mean(axis=axis)
def test_sum_and_mean(self):
for fn in [ch.sum, ch.mean]:
data = ch.zeros((3,4,7,2))
dsum = fn(data, axis=2)
dr = dsum.dr_wrt(data)
diff = ch.random.randn(data.size).reshape(data.shape)
pred = dr.dot(diff.r.ravel())
gt = fn(diff, axis=2)
#print pred
#print gt
#print pred.ravel() - gt.r.ravel()
self.assertTrue(1e-15 > np.max(np.abs(gt.r.ravel() - pred)))
# test caching
dr0 = gt.dr_wrt(diff)
diff[:] = np.random.randn(diff.size).reshape(diff.shape)
self.assertTrue(gt.dr_wrt(diff) is dr0) # changing values shouldn't force recompute
gt.axis=1
self.assertTrue(gt.dr_wrt(diff) is not dr0)
