def get_id_minus_conjugate_z_times_z(z: torch.Tensor):
"""
:param z: b x 2 x n x n
:return: Id - \overline(z)z
"""
identity = sm.identity_like(z)
conj_z_z = sm.bmm(sm.conjugate(z), z)
return sm.subtract(identity, conj_z_z)
def test_takagi_factorization_imag_identity(self):
a = sm.identity_like(get_random_symmetric_matrices(3, 3))
a = sm.stick(sm.imag(a), sm.real(a))
eigenvalues, s = TakagiFactorization(3).factorize(a)
diagonal = torch.diag_embed(eigenvalues)
diagonal = sm.stick(diagonal, torch.zeros_like(diagonal))
self.assertAllClose(
a, sm.bmm3(sm.conjugate(s), diagonal, sm.conj_trans(s)))
def inverse_cayley_transform(z: torch.Tensor) -> torch.Tensor:
"""
T^-1(Z): Bounded Domain Model -> Upper Half Space model
T^-1(Z) = i (Id + Z)(Id - Z)^-1
:param z: b x 2 x n x n: PRE: z \in Bounded Domain Manifold
:return: y \in Upper Half Space Manifold
"""
identity = sm.identity_like(z)
i_z_plus_id = sm.multiply_by_i(sm.add(identity, z))
inv_z_minus_id = sm.inverse(sm.subtract(identity, z))
return sm.bmm(i_z_plus_id, inv_z_minus_id)
def cayley_transform(z: torch.Tensor) -> torch.Tensor:
"""
T(Z): Upper Half Space model -> Bounded Domain Model
T(Z) = (Z - i Id)(Z + i Id)^-1
:param z: b x 2 x n x n: PRE: z \in Upper Half Space Manifold
:return: y \in Bounded Domain Manifold
"""
identity = sm.identity_like(z)
i_identity = sm.stick(sm.imag(identity), sm.real(identity))
z_minus_id = sm.subtract(z, i_identity)
inv_z_plus_id = sm.inverse(sm.add(z, i_identity))
return sm.bmm(z_minus_id, inv_z_plus_id)
def test_takagi_factorization_real_diagonal(self):
a = get_random_symmetric_matrices(3, 3) * 10
a = torch.where(sm.identity_like(a) == 1, a, torch.zeros_like(a))
eigenvalues, s = TakagiFactorization(3).factorize(a)
diagonal = torch.diag_embed(eigenvalues)
diagonal = sm.stick(diagonal, torch.zeros_like(diagonal))
self.assertAllClose(
a, sm.bmm3(sm.conjugate(s), diagonal, sm.conj_trans(s)))
# real part of eigenvectors is made of vectors with one 1 and all zeros
real_part = torch.sum(torch.abs(sm.real(s)), dim=-1)
self.assertAllClose(torch.ones_like(real_part), real_part)
# imaginary part of eigenvectors is all zeros
self.assertAllClose(torch.zeros(1), torch.sum(sm.imag(s)))
def egrad2rgrad(self, z: torch.Tensor, u: torch.Tensor) -> torch.Tensor:
"""
Transform gradient computed using autodiff to the correct Riemannian gradient for the point :math:`x`.
If you have a function f(z) on Mn, then the Riemannian gradient is
grad_R(f(z)) = (Id + ẑz) * grad_E(f(z)) * (Id + zẑ)
:param z: point on the manifold. Shape: (b, 2, n, n)
:param u: gradient to be projected: Shape: same than z
:return grad vector in the Riemannian manifold. Shape: same than z
"""
id = sm.identity_like(z)
conjz = sm.conjugate(z)
id_plus_conjz_z = id + sm.bmm(conjz, z)
id_plus_z_conjz = id + sm.bmm(z, conjz)
riem_grad = sm.bmm3(id_plus_conjz_z, u, id_plus_z_conjz)
return riem_grad
def inner(self,
z: torch.Tensor,
u: torch.Tensor,
v=None,
*,
keepdim=False) -> torch.Tensor:
"""
Inner product for tangent vectors at point :math:`z`.
For the bounded domain model, the inner product at point z of the vectors u, v
it is (z, u, v are complex symmetric matrices):
g_{z}(u, v) = tr[ (Id - ẑz)^-1 u (Id - zẑ)^-1 \ov{v} ]
:param z: torch.Tensor point on the manifold: b x 2 x n x n
:param u: torch.Tensor tangent vector at point :math:`z`: b x 2 x n x n
:param v: Optional[torch.Tensor] tangent vector at point :math:`z`: b x 2 x n x n
:param keepdim: bool keep the last dim?
:return: torch.Tensor inner product (broadcastable): b x 2 x 1 x 1
"""
if v is None:
v = u
identity = sm.identity_like(z)
conj_z = sm.conjugate(z)
conj_z_z = sm.bmm(conj_z, z)
z_conj_z = sm.bmm(z, conj_z)
inv_id_minus_conj_z_z = sm.subtract(identity, conj_z_z)
inv_id_minus_z_conj_z = sm.subtract(identity, z_conj_z)
inv_id_minus_conj_z_z = sm.inverse(inv_id_minus_conj_z_z)
inv_id_minus_z_conj_z = sm.inverse(inv_id_minus_z_conj_z)
res = sm.bmm3(inv_id_minus_conj_z_z, u, inv_id_minus_z_conj_z)
res = sm.bmm(res, sm.conjugate(v))
real_part = sm.trace(sm.real(res), keepdim=True)
real_part = torch.unsqueeze(real_part, -1) # b x 1 x 1
return sm.stick(real_part, real_part) # # b x 2 x 1 x 1