def do(self, a, b):
d = linalg.det(a)
(s, ld) = linalg.slogdet(a)
if asarray(a).dtype.type in (single, double):
ad = asarray(a).astype(double)
else:
ad = asarray(a).astype(cdouble)
ev = linalg.eigvals(ad)
assert_almost_equal(d, multiply.reduce(ev, axis=-1))
assert_almost_equal(s * np.exp(ld), multiply.reduce(ev, axis=-1))
s = np.atleast_1d(s)
ld = np.atleast_1d(ld)
m = (s != 0)
assert_almost_equal(np.abs(s[m]), 1)
assert_equal(ld[~m], -inf)
python类slogdet()的实例源码
def do(self, a, b):
d = linalg.det(a)
(s, ld) = linalg.slogdet(a)
if asarray(a).dtype.type in (single, double):
ad = asarray(a).astype(double)
else:
ad = asarray(a).astype(cdouble)
ev = linalg.eigvals(ad)
assert_almost_equal(d, multiply.reduce(ev, axis=-1))
assert_almost_equal(s * np.exp(ld), multiply.reduce(ev, axis=-1))
s = np.atleast_1d(s)
ld = np.atleast_1d(ld)
m = (s != 0)
assert_almost_equal(np.abs(s[m]), 1)
assert_equal(ld[~m], -inf)
def log_marg_k(self, k):
"""
Return the log marginal probability of the data vectors assigned to
component `k`.
The log marginal probability p(X) = p(x_1, x_2, ..., x_N) is returned
for the data vectors assigned to component `k`. See (266) in Murphy's
bayesGauss notes, p. 21.
"""
k_N = self.prior.k_0 + self.counts[k]
v_N = self.prior.v_0 + self.counts[k]
m_N = self.m_N_numerators[k]/k_N
S_N = self.S_N_partials[k] - k_N*np.outer(m_N, m_N)
i = np.arange(1, self.D + 1, dtype=np.int)
return (
- self.counts[k]*self.D/2.*self._cached_log_pi
+ self.D/2.*math.log(self.prior.k_0) - self.D/2.*math.log(k_N)
+ self.prior.v_0/2.*slogdet(self.prior.S_0)[1]
- v_N/2.*slogdet(S_N)[1]
+ np.sum(
self._cached_gammaln_by_2[v_N + 1 - i] -
self._cached_gammaln_by_2[self.prior.v_0 + 1 - i]
)
)
test_linalg.py 文件源码
项目：PyDataLondon29-EmbarrassinglyParallelDAWithAWSLambda
作者: SignalMedia
项目源码
文件源码
阅读 23
收藏 0
点赞 0
评论 0
def do(self, a, b):
d = linalg.det(a)
(s, ld) = linalg.slogdet(a)
if asarray(a).dtype.type in (single, double):
ad = asarray(a).astype(double)
else:
ad = asarray(a).astype(cdouble)
ev = linalg.eigvals(ad)
assert_almost_equal(d, multiply.reduce(ev, axis=-1))
assert_almost_equal(s * np.exp(ld), multiply.reduce(ev, axis=-1))
s = np.atleast_1d(s)
ld = np.atleast_1d(ld)
m = (s != 0)
assert_almost_equal(np.abs(s[m]), 1)
assert_equal(ld[~m], -inf)
def do(self, a, b):
d = linalg.det(a)
(s, ld) = linalg.slogdet(a)
if asarray(a).dtype.type in (single, double):
ad = asarray(a).astype(double)
else:
ad = asarray(a).astype(cdouble)
ev = linalg.eigvals(ad)
assert_almost_equal(d, multiply.reduce(ev, axis=-1))
assert_almost_equal(s * np.exp(ld), multiply.reduce(ev, axis=-1))
s = np.atleast_1d(s)
ld = np.atleast_1d(ld)
m = (s != 0)
assert_almost_equal(np.abs(s[m]), 1)
assert_equal(ld[~m], -inf)
def do(self, a, b):
d = linalg.det(a)
(s, ld) = linalg.slogdet(a)
if asarray(a).dtype.type in (single, double):
ad = asarray(a).astype(double)
else:
ad = asarray(a).astype(cdouble)
ev = linalg.eigvals(ad)
assert_almost_equal(d, multiply.reduce(ev, axis=-1))
assert_almost_equal(s * np.exp(ld), multiply.reduce(ev, axis=-1))
s = np.atleast_1d(s)
ld = np.atleast_1d(ld)
m = (s != 0)
assert_almost_equal(np.abs(s[m]), 1)
assert_equal(ld[~m], -inf)
def do(self, a, b):
d = linalg.det(a)
(s, ld) = linalg.slogdet(a)
if asarray(a).dtype.type in (single, double):
ad = asarray(a).astype(double)
else:
ad = asarray(a).astype(cdouble)
ev = linalg.eigvals(ad)
assert_almost_equal(d, multiply.reduce(ev, axis=-1))
assert_almost_equal(s * np.exp(ld), multiply.reduce(ev, axis=-1))
s = np.atleast_1d(s)
ld = np.atleast_1d(ld)
m = (s != 0)
assert_almost_equal(np.abs(s[m]), 1)
assert_equal(ld[~m], -inf)
def log_marg_k(self, k):
"""
Return the log marginal probability of the data vectors assigned to
component `k`.
The log marginal probability p(X) = p(x_1, x_2, ..., x_N) is returned
for the data vectors assigned to component `k`. See (266) in Murphy's
bayesGauss notes, p. 21.
"""
k_N = self.prior.k_0 + self.counts[k]
v_N = self.prior.v_0 + self.counts[k]
m_N = self.m_N_numerators[k]/k_N
S_N = self.S_N_partials[k] - k_N*np.outer(m_N, m_N)
i = np.arange(1, self.D + 1, dtype=np.int)
return (
- self.counts[k]*self.D/2.*self._cached_log_pi
+ self.D/2.*math.log(self.prior.k_0) - self.D/2.*math.log(k_N)
+ self.prior.v_0/2.*slogdet(self.prior.S_0)[1]
- v_N/2.*slogdet(S_N)[1]
+ np.sum(
self._cached_gammaln_by_2[v_N + 1 - i] -
self._cached_gammaln_by_2[self.prior.v_0 + 1 - i]
)
)
def do(self, a, b):
d = linalg.det(a)
(s, ld) = linalg.slogdet(a)
if asarray(a).dtype.type in (single, double):
ad = asarray(a).astype(double)
else:
ad = asarray(a).astype(cdouble)
ev = linalg.eigvals(ad)
assert_almost_equal(d, multiply.reduce(ev, axis=-1))
assert_almost_equal(s * np.exp(ld), multiply.reduce(ev, axis=-1))
s = np.atleast_1d(s)
ld = np.atleast_1d(ld)
m = (s != 0)
assert_almost_equal(np.abs(s[m]), 1)
assert_equal(ld[~m], -inf)
def test_zero(self):
assert_equal(linalg.det([[0.0]]), 0.0)
assert_equal(type(linalg.det([[0.0]])), double)
assert_equal(linalg.det([[0.0j]]), 0.0)
assert_equal(type(linalg.det([[0.0j]])), cdouble)
assert_equal(linalg.slogdet([[0.0]]), (0.0, -inf))
assert_equal(type(linalg.slogdet([[0.0]])[0]), double)
assert_equal(type(linalg.slogdet([[0.0]])[1]), double)
assert_equal(linalg.slogdet([[0.0j]]), (0.0j, -inf))
assert_equal(type(linalg.slogdet([[0.0j]])[0]), cdouble)
assert_equal(type(linalg.slogdet([[0.0j]])[1]), double)
def test_types(self):
def check(dtype):
x = np.array([[1, 0.5], [0.5, 1]], dtype=dtype)
assert_equal(np.linalg.det(x).dtype, dtype)
ph, s = np.linalg.slogdet(x)
assert_equal(s.dtype, get_real_dtype(dtype))
assert_equal(ph.dtype, dtype)
for dtype in [single, double, csingle, cdouble]:
yield check, dtype
def test_zero(self):
assert_equal(linalg.det([[0.0]]), 0.0)
assert_equal(type(linalg.det([[0.0]])), double)
assert_equal(linalg.det([[0.0j]]), 0.0)
assert_equal(type(linalg.det([[0.0j]])), cdouble)
assert_equal(linalg.slogdet([[0.0]]), (0.0, -inf))
assert_equal(type(linalg.slogdet([[0.0]])[0]), double)
assert_equal(type(linalg.slogdet([[0.0]])[1]), double)
assert_equal(linalg.slogdet([[0.0j]]), (0.0j, -inf))
assert_equal(type(linalg.slogdet([[0.0j]])[0]), cdouble)
assert_equal(type(linalg.slogdet([[0.0j]])[1]), double)
def test_types(self):
def check(dtype):
x = np.array([[1, 0.5], [0.5, 1]], dtype=dtype)
assert_equal(np.linalg.det(x).dtype, dtype)
ph, s = np.linalg.slogdet(x)
assert_equal(s.dtype, get_real_dtype(dtype))
assert_equal(ph.dtype, dtype)
for dtype in [single, double, csingle, cdouble]:
yield check, dtype
def log_prior(self, i):
"""Return the probability of `X[i]` under the prior alone."""
mu = self.prior.m_0
covar = (self.prior.k_0 + 1) / (self.prior.k_0*(self.prior.v_0 - self.D + 1)) * self.prior.S_0
logdet_covar = slogdet(covar)[1]
inv_covar = inv(covar)
v = self.prior.v_0 - self.D + 1
return self._multivariate_students_t(i, mu, logdet_covar, inv_covar, v)
def _update_logdet_covar_and_inv_covar(self, k):
"""
Update the covariance terms for component `k`.
Based on the `m_N_numerators` and `S_N_partials` terms for the `k`th
component, the `logdet_covars` and `inv_covars` terms are updated.
"""
k_N = self.prior.k_0 + self.counts[k]
v_N = self.prior.v_0 + self.counts[k]
m_N = self.m_N_numerators[k]/k_N
covar = (k_N + 1.)/(k_N*(v_N - self.D + 1.)) * (self.S_N_partials[k] - k_N*np.outer(m_N, m_N))
self.logdet_covars[k] = slogdet(covar)[1]
self.inv_covars[k, :, :] = inv(covar)
# @profile
def logpdf(x, df, mu, Sigma):
"""
Marginal log-likelihood of a Student-t Process
Parameters
----------
x: array-like
Point to be evaluated
df: float
Degrees of freedom (>2.0)
mu: array-like
Mean of the process.
Sigma: array-like
Covariance matrix of the process.
Returns
-------
logp: float
log-likelihood
"""
d = len(x)
x = np.atleast_2d(x)
xm = x - mu
V = df * Sigma
V_inv = np.linalg.inv(V)
_, logdet = slogdet(np.pi * V)
logz = -gamma(df / 2.0 + d / 2.0) + gamma(df / 2.0) + 0.5 * logdet
logp = -0.5 * (df + d) * np.log(1 + np.sum(np.dot(xm, V_inv) * xm, axis=1))
logp = logp - logz
return logp[0]
test_linalg.py 文件源码
项目：PyDataLondon29-EmbarrassinglyParallelDAWithAWSLambda
作者: SignalMedia
项目源码
文件源码
阅读 25
收藏 0
点赞 0
评论 0
def test_zero(self):
assert_equal(linalg.det([[0.0]]), 0.0)
assert_equal(type(linalg.det([[0.0]])), double)
assert_equal(linalg.det([[0.0j]]), 0.0)
assert_equal(type(linalg.det([[0.0j]])), cdouble)
assert_equal(linalg.slogdet([[0.0]]), (0.0, -inf))
assert_equal(type(linalg.slogdet([[0.0]])[0]), double)
assert_equal(type(linalg.slogdet([[0.0]])[1]), double)
assert_equal(linalg.slogdet([[0.0j]]), (0.0j, -inf))
assert_equal(type(linalg.slogdet([[0.0j]])[0]), cdouble)
assert_equal(type(linalg.slogdet([[0.0j]])[1]), double)
test_linalg.py 文件源码
项目：PyDataLondon29-EmbarrassinglyParallelDAWithAWSLambda
作者: SignalMedia
项目源码
文件源码
阅读 23
收藏 0
点赞 0
评论 0
def test_types(self):
def check(dtype):
x = np.array([[1, 0.5], [0.5, 1]], dtype=dtype)
assert_equal(np.linalg.det(x).dtype, dtype)
ph, s = np.linalg.slogdet(x)
assert_equal(s.dtype, get_real_dtype(dtype))
assert_equal(ph.dtype, dtype)
for dtype in [single, double, csingle, cdouble]:
yield check, dtype
def test_zero(self):
assert_equal(linalg.det([[0.0]]), 0.0)
assert_equal(type(linalg.det([[0.0]])), double)
assert_equal(linalg.det([[0.0j]]), 0.0)
assert_equal(type(linalg.det([[0.0j]])), cdouble)
assert_equal(linalg.slogdet([[0.0]]), (0.0, -inf))
assert_equal(type(linalg.slogdet([[0.0]])[0]), double)
assert_equal(type(linalg.slogdet([[0.0]])[1]), double)
assert_equal(linalg.slogdet([[0.0j]]), (0.0j, -inf))
assert_equal(type(linalg.slogdet([[0.0j]])[0]), cdouble)
assert_equal(type(linalg.slogdet([[0.0j]])[1]), double)
def test_types(self):
def check(dtype):
x = np.array([[1, 0.5], [0.5, 1]], dtype=dtype)
assert_equal(np.linalg.det(x).dtype, dtype)
ph, s = np.linalg.slogdet(x)
assert_equal(s.dtype, get_real_dtype(dtype))
assert_equal(ph.dtype, dtype)
for dtype in [single, double, csingle, cdouble]:
yield check, dtype
def test_zero(self):
assert_equal(linalg.det([[0.0]]), 0.0)
assert_equal(type(linalg.det([[0.0]])), double)
assert_equal(linalg.det([[0.0j]]), 0.0)
assert_equal(type(linalg.det([[0.0j]])), cdouble)
assert_equal(linalg.slogdet([[0.0]]), (0.0, -inf))
assert_equal(type(linalg.slogdet([[0.0]])[0]), double)
assert_equal(type(linalg.slogdet([[0.0]])[1]), double)
assert_equal(linalg.slogdet([[0.0j]]), (0.0j, -inf))
assert_equal(type(linalg.slogdet([[0.0j]])[0]), cdouble)
assert_equal(type(linalg.slogdet([[0.0j]])[1]), double)
def test_types(self):
def check(dtype):
x = np.array([[1, 0.5], [0.5, 1]], dtype=dtype)
assert_equal(np.linalg.det(x).dtype, dtype)
ph, s = np.linalg.slogdet(x)
assert_equal(s.dtype, get_real_dtype(dtype))
assert_equal(ph.dtype, dtype)
for dtype in [single, double, csingle, cdouble]:
yield check, dtype
def test_zero(self):
assert_equal(linalg.det([[0.0]]), 0.0)
assert_equal(type(linalg.det([[0.0]])), double)
assert_equal(linalg.det([[0.0j]]), 0.0)
assert_equal(type(linalg.det([[0.0j]])), cdouble)
assert_equal(linalg.slogdet([[0.0]]), (0.0, -inf))
assert_equal(type(linalg.slogdet([[0.0]])[0]), double)
assert_equal(type(linalg.slogdet([[0.0]])[1]), double)
assert_equal(linalg.slogdet([[0.0j]]), (0.0j, -inf))
assert_equal(type(linalg.slogdet([[0.0j]])[0]), cdouble)
assert_equal(type(linalg.slogdet([[0.0j]])[1]), double)
def test_types(self):
def check(dtype):
x = np.array([[1, 0.5], [0.5, 1]], dtype=dtype)
assert_equal(np.linalg.det(x).dtype, dtype)
ph, s = np.linalg.slogdet(x)
assert_equal(s.dtype, get_real_dtype(dtype))
assert_equal(ph.dtype, dtype)
for dtype in [single, double, csingle, cdouble]:
yield check, dtype
def log_prior(self, i):
"""Return the probability of `X[i]` under the prior alone."""
mu = self.prior.m_0
covar = (self.prior.k_0 + 1) / (self.prior.k_0*(self.prior.v_0 - self.D + 1)) * self.prior.S_0
logdet_covar = slogdet(covar)[1]
inv_covar = inv(covar)
v = self.prior.v_0 - self.D + 1
return self._multivariate_students_t(i, mu, logdet_covar, inv_covar, v)
def _update_logdet_covar_and_inv_covar(self, k):
"""
Update the covariance terms for component `k`.
Based on the `m_N_numerators` and `S_N_partials` terms for the `k`th
component, the `logdet_covars` and `inv_covars` terms are updated.
"""
k_N = self.prior.k_0 + self.counts[k]
v_N = self.prior.v_0 + self.counts[k]
m_N = self.m_N_numerators[k]/k_N
covar = (k_N + 1.)/(k_N*(v_N - self.D + 1.)) * (self.S_N_partials[k] - k_N*np.outer(m_N, m_N))
self.logdet_covars[k] = slogdet(covar)[1]
self.inv_covars[k, :, :] = inv(covar)
# @profile
def test_zero(self):
assert_equal(linalg.det([[0.0]]), 0.0)
assert_equal(type(linalg.det([[0.0]])), double)
assert_equal(linalg.det([[0.0j]]), 0.0)
assert_equal(type(linalg.det([[0.0j]])), cdouble)
assert_equal(linalg.slogdet([[0.0]]), (0.0, -inf))
assert_equal(type(linalg.slogdet([[0.0]])[0]), double)
assert_equal(type(linalg.slogdet([[0.0]])[1]), double)
assert_equal(linalg.slogdet([[0.0j]]), (0.0j, -inf))
assert_equal(type(linalg.slogdet([[0.0j]])[0]), cdouble)
assert_equal(type(linalg.slogdet([[0.0j]])[1]), double)
def test_types(self):
def check(dtype):
x = np.array([[1, 0.5], [0.5, 1]], dtype=dtype)
assert_equal(np.linalg.det(x).dtype, dtype)
ph, s = np.linalg.slogdet(x)
assert_equal(s.dtype, get_real_dtype(dtype))
assert_equal(ph.dtype, dtype)
for dtype in [single, double, csingle, cdouble]:
yield check, dtype
