def logdet(self):
d, s = self._d, self._s_perp
fsh = full_shape([d.shape, s.shape])
logdet = np.zeros(fsh)
for idx in itertools.product(*map(range, fsh)):
d_idx, s_idx = broadcast(idx, d.shape), broadcast(idx, s.shape)
logdet[idx] = np.log(d[d_idx]).sum() + \
(self.dim - d[d_idx].size) * np.log(s[s_idx])
return logdet
def __init__(self, mu, Sigma, Z=0.):
mu, Sigma, Z = match_shapes([('mu', mu, 1), ('Sigma', Sigma, 0),
('Z', Z, 0)])
self._mu = mu
self._Sigma = Sigma
self._Z = Z
self.dim = mu.shape[-1]
self.ndim = mu.ndim - 1
self.shape = full_shape([Sigma.shape, mu.shape[:-1], Z.shape])
self.shape_str = '%s mu=%s Z=%s %s' % (Sigma.__class__, mu.shape,
Z.shape, Sigma.shape_str)
def __init__(self, d, Q, s_perp, dim):
BaseMatrix.__init__(self)
d, Q, s_perp = match_shapes([('d', d, 0), ('Q', Q, 0),
('s_perp', s_perp, 0)])
self._d = d
self._Q = Q
self.dim = dim
self.ndim = d.ndim
self._s_perp = s_perp
self.shape = full_shape([d.shape, Q.shape, s_perp.shape])
self.shape_str = 'd=%s Q=%s s_perp=%s' % (d.shape, Q.shape,
s_perp.shape)
def _QDQ_x(Q, d, x):
fsh = full_shape([x.shape[:-1], Q.shape, d.shape])
prod = np.zeros(fsh + (x.shape[-1], ))
for full_idx in itertools.product(*map(range, fsh)):
Q_idx = broadcast(full_idx, Q.shape)
d_idx = broadcast(full_idx, d.shape)
x_idx = broadcast(full_idx, x.shape[:-1])
curr_d, curr_Q = d[d_idx], Q[Q_idx]
curr_x = x[x_idx + (slice(None), )]
curr_QTx = np.dot(curr_Q.T, curr_x)
prod[full_idx + (slice(None), )] = np.dot(curr_Q, curr_d * curr_QTx)
return prod
def __init__(self, J, Lambda, Z):
J, Lambda, Z = match_shapes([('J', J, 1), ('Lambda', Lambda, 0),
('Z', Z, 0)])
self._J = J
self._Lambda = Lambda
self._Z = Z
self.shape = full_shape([J.shape[:-1], Lambda.shape, Z.shape])
self.ndim = J.ndim - 1
self.dim = J.shape[-1]
self.shape_str = '%s J=%s Z=%s %s' % (Lambda.__class__, J.shape,
Z.shape, Lambda.shape_str)
self.mutable = False
def __init__(self, d, Q, s_perp):
BaseMatrix.__init__(self)
d, Q, s_perp = match_shapes([('d', d, 1), ('Q', Q, 2),
('s_perp', s_perp, 0)])
self._d = d
self._Q = Q
self.dim = Q.shape[-2]
self.rank = Q.shape[-1]
self.ndim = Q.ndim - 2
self.shape = full_shape([d.shape[:-1], Q.shape[:-2], s_perp.shape])
self._s_perp = s_perp
self.shape_str = 'd=%s Q=%s s_perp=%s' % (d.shape, Q.shape,
s_perp.shape)
def full(self):
S = np.zeros(
full_shape([self._d.shape, self._Q.shape]) + (self.dim, self.dim))
for idx in itertools.product(*map(range, S.shape[:-2])):
d_idx, Q_idx = broadcast(idx, self._d.shape), broadcast(
idx, self._Q.shape)
d, Q = self._d[d_idx], self._Q[Q_idx]
S[idx + (Ellipsis, )] = np.dot(Q * d, Q.T)
sp_idx = broadcast(idx, self._s_perp.shape)
sp = self._s_perp[sp_idx]
S[idx + (Ellipsis, )] += (np.eye(self.dim) - np.dot(Q, Q.T)) * sp
return FullMatrix(S)
def __init__(self, Lambda, J_diff, Z_diff, X):
Lambda, J_diff, X = match_shapes([('Lambda', Lambda, 0),
('J_diff', J_diff, 1), ('X', X, 1)])
self._Lambda = Lambda
self._J_diff = J_diff.copy()
self._Z_diff = Z_diff.copy()
self._X = X.copy()
self.dim = self._J_diff.shape[-1]
self.ndim = self._J_diff.ndim - 1
self.shape = full_shape(
[Lambda.shape, J_diff.shape[:-1], X.shape[:-1]])
self.shape_str = '%s J_diff=%s X=%s %s' % (
Lambda.__class__, J_diff.shape, X.shape, Lambda.shape_str)