]> gitweb.michael.orlitzky.com - sage.d.git/commitdiff
Remove the Pi/Z stuff for inclusion into sage.
authorMichael Orlitzky <michael@orlitzky.com>
Sun, 25 Sep 2016 19:41:41 +0000 (15:41 -0400)
committerMichael Orlitzky <michael@orlitzky.com>
Sun, 25 Sep 2016 19:41:41 +0000 (15:41 -0400)
mjo/cone/cone.py

index fd63612b3277151d880be83fa5b827652081bd22..7e9c549eec66ede6dc0c94bd67e0e16d4d999538 100644 (file)
@@ -70,746 +70,24 @@ def is_lyapunov_like(L,K):
     return all([(L*x).inner_product(s) == 0
                 for (x,s) in K.discrete_complementarity_set()])
 
-
-def positive_operator_gens(K1, K2 = None):
-    r"""
-    Compute generators of the cone of positive operators on this cone. A
-    linear operator on a cone is positive if the image of the cone under
-    the operator is a subset of the cone. This concept can be extended
-    to two cones, where the image of the first cone under a positive
-    operator is a subset of the second cone.
-
-    INPUT:
-
-    - ``K2`` -- (default: ``K1``) the codomain cone; the image of this
-                cone under the returned operators is a subset of ``K2``.
-
-    OUTPUT:
-
-    A list of `m`-by-``n`` matrices where ``m == K2.lattice_dim()`` and
-    ``n == K1.lattice_dim()``. Each matrix ``P`` in the list should have
-    the property that ``P*x`` is an element of ``K2`` whenever ``x`` is
-    an element of ``K1``. Moreover, any nonnegative linear combination of
-    these matrices shares the same property.
-
-    .. SEEALSO::
-
-           :meth:`cross_positive_operator_gens`, :meth:`Z_operator_gens`,
-
-    REFERENCES:
-
-    .. [Tam]
-       B.-S. Tam.
-       Some results of polyhedral cones and simplicial cones.
-       Linear and Multilinear Algebra, 4:4 (1977) 281--284.
-
-    EXAMPLES:
-
-    Positive operators on the nonnegative orthant are nonnegative matrices::
-
-        sage: K = Cone([(1,)])
-        sage: positive_operator_gens(K)
-        [[1]]
-
-        sage: K = Cone([(1,0),(0,1)])
-        sage: positive_operator_gens(K)
-        [
-        [1 0]  [0 1]  [0 0]  [0 0]
-        [0 0], [0 0], [1 0], [0 1]
-        ]
-
-    The trivial cone in a trivial space has no positive operators::
-
-        sage: K = Cone([], ToricLattice(0))
-        sage: positive_operator_gens(K)
-        []
-
-    Every operator is positive on the trivial cone::
-
-        sage: K = Cone([(0,)])
-        sage: positive_operator_gens(K)
-        [[1], [-1]]
-
-        sage: K = Cone([(0,0)])
-        sage: K.is_trivial()
-        True
-        sage: positive_operator_gens(K)
-        [
-        [1 0]  [-1  0]  [0 1]  [ 0 -1]  [0 0]  [ 0  0]  [0 0]  [ 0  0]
-        [0 0], [ 0  0], [0 0], [ 0  0], [1 0], [-1  0], [0 1], [ 0 -1]
-        ]
-
-    Every operator is positive on the ambient vector space::
-
-        sage: K = Cone([(1,),(-1,)])
-        sage: K.is_full_space()
-        True
-        sage: positive_operator_gens(K)
-        [[1], [-1]]
-
-        sage: K = Cone([(1,0),(-1,0),(0,1),(0,-1)])
-        sage: K.is_full_space()
-        True
-        sage: positive_operator_gens(K)
-        [
-        [1 0]  [-1  0]  [0 1]  [ 0 -1]  [0 0]  [ 0  0]  [0 0]  [ 0  0]
-        [0 0], [ 0  0], [0 0], [ 0  0], [1 0], [-1  0], [0 1], [ 0 -1]
-        ]
-
-    A non-obvious application is to find the positive operators on the
-    right half-plane::
-
-        sage: K = Cone([(1,0),(0,1),(0,-1)])
-        sage: positive_operator_gens(K)
-        [
-        [1 0]  [0 0]  [ 0  0]  [0 0]  [ 0  0]
-        [0 0], [1 0], [-1  0], [0 1], [ 0 -1]
-        ]
-
-    TESTS:
-
-    Each positive operator generator should send the generators of one
-    cone into the other cone::
-
-        sage: set_random_seed()
-        sage: K1 = random_cone(max_ambient_dim=4)
-        sage: K2 = random_cone(max_ambient_dim=4)
-        sage: pi_K1_K2 = positive_operator_gens(K1,K2)
-        sage: all([ K2.contains(P*x) for P in pi_K1_K2 for x in K1 ])
-        True
-
-    Each positive operator generator should send a random element of one
-    cone into the other cone::
-
-        sage: set_random_seed()
-        sage: K1 = random_cone(max_ambient_dim=4)
-        sage: K2 = random_cone(max_ambient_dim=4)
-        sage: pi_K1_K2 = positive_operator_gens(K1,K2)
-        sage: all([ K2.contains(P*K1.random_element(QQ)) for P in pi_K1_K2 ])
-        True
-
-    A random element of the positive operator cone should send the
-    generators of one cone into the other cone::
-
-        sage: set_random_seed()
-        sage: K1 = random_cone(max_ambient_dim=4)
-        sage: K2 = random_cone(max_ambient_dim=4)
-        sage: pi_K1_K2 = positive_operator_gens(K1,K2)
-        sage: L = ToricLattice(K1.lattice_dim() * K2.lattice_dim())
-        sage: pi_cone = Cone([ g.list() for g in pi_K1_K2 ],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: P = matrix(K2.lattice_dim(),
-        ....:            K1.lattice_dim(),
-        ....:            pi_cone.random_element(QQ).list())
-        sage: all([ K2.contains(P*x) for x in K1 ])
-        True
-
-    A random element of the positive operator cone should send a random
-    element of one cone into the other cone::
-
-        sage: set_random_seed()
-        sage: K1 = random_cone(max_ambient_dim=4)
-        sage: K2 = random_cone(max_ambient_dim=4)
-        sage: pi_K1_K2 = positive_operator_gens(K1,K2)
-        sage: L = ToricLattice(K1.lattice_dim() * K2.lattice_dim())
-        sage: pi_cone = Cone([ g.list() for g in pi_K1_K2 ],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: P = matrix(K2.lattice_dim(),
-        ....:            K1.lattice_dim(),
-        ....:            pi_cone.random_element(QQ).list())
-        sage: K2.contains(P*K1.random_element(ring=QQ))
-        True
-
-    The lineality space of the dual of the cone of positive operators
-    can be computed from the lineality spaces of the cone and its dual::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: L = ToricLattice(K.lattice_dim()**2)
-        sage: pi_cone = Cone([ g.list() for g in pi_of_K ],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: actual = pi_cone.dual().linear_subspace()
-        sage: U1 = [ vector((s.tensor_product(x)).list())
-        ....:        for x in K.lines()
-        ....:        for s in K.dual() ]
-        sage: U2 = [ vector((s.tensor_product(x)).list())
-        ....:        for x in K
-        ....:        for s in K.dual().lines() ]
-        sage: expected = pi_cone.lattice().vector_space().span(U1 + U2)
-        sage: actual == expected
-        True
-
-    The lineality of the dual of the cone of positive operators
-    is known from its lineality space::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: n = K.lattice_dim()
-        sage: m = K.dim()
-        sage: l = K.lineality()
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: L = ToricLattice(n**2)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                 lattice=L,
-        ....:                 check=False)
-        sage: actual = pi_cone.dual().lineality()
-        sage: expected = l*(m - l) + m*(n - m)
-        sage: actual == expected
-        True
-
-    The dimension of the cone of positive operators is given by the
-    corollary in my paper::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: n = K.lattice_dim()
-        sage: m = K.dim()
-        sage: l = K.lineality()
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: L = ToricLattice(n**2)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: actual = pi_cone.dim()
-        sage: expected = n**2 - l*(m - l) - (n - m)*m
-        sage: actual == expected
-        True
-
-    The trivial cone, full space, and half-plane all give rise to the
-    expected dimensions::
-
-        sage: n = ZZ.random_element().abs()
-        sage: K = Cone([[0] * n], ToricLattice(n))
-        sage: K.is_trivial()
-        True
-        sage: L = ToricLattice(n^2)
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: actual = pi_cone.dim()
-        sage: actual == n^2
-        True
-        sage: K = K.dual()
-        sage: K.is_full_space()
-        True
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: actual = pi_cone.dim()
-        sage: actual == n^2
-        True
-        sage: K = Cone([(1,0),(0,1),(0,-1)])
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: actual = Cone([p.list() for p in pi_of_K], check=False).dim()
-        sage: actual == 3
-        True
-
-    The lineality of the cone of positive operators follows from the
-    description of its generators::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: n = K.lattice_dim()
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: L = ToricLattice(n**2)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: actual = pi_cone.lineality()
-        sage: expected = n**2 - K.dim()*K.dual().dim()
-        sage: actual == expected
-        True
-
-    The trivial cone, full space, and half-plane all give rise to the
-    expected linealities::
-
-        sage: n = ZZ.random_element().abs()
-        sage: K = Cone([[0] * n], ToricLattice(n))
-        sage: K.is_trivial()
-        True
-        sage: L = ToricLattice(n^2)
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: actual = pi_cone.lineality()
-        sage: actual == n^2
-        True
-        sage: K = K.dual()
-        sage: K.is_full_space()
-        True
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K], lattice=L)
-        sage: pi_cone.lineality() == n^2
-        True
-        sage: K = Cone([(1,0),(0,1),(0,-1)])
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K], check=False)
-        sage: actual = pi_cone.lineality()
-        sage: actual == 2
-        True
-
-    A cone is proper if and only if its cone of positive operators
-    is proper::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: L = ToricLattice(K.lattice_dim()**2)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: K.is_proper() == pi_cone.is_proper()
-        True
-
-    The positive operators of a permuted cone can be obtained by
-    conjugation::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: L = ToricLattice(K.lattice_dim()**2)
-        sage: p = SymmetricGroup(K.lattice_dim()).random_element().matrix()
-        sage: pK = Cone([ p*k for k in K ], K.lattice(), check=False)
-        sage: pi_of_pK = positive_operator_gens(pK)
-        sage: actual = Cone([t.list() for t in pi_of_pK],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: expected = Cone([(p*t*p.inverse()).list() for t in pi_of_K],
-        ....:                   lattice=L,
-        ....:                   check=False)
-        sage: actual.is_equivalent(expected)
-        True
-
-    A transformation is positive on a cone if and only if its adjoint is
-    positive on the dual of that cone::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: F = K.lattice().vector_space().base_field()
-        sage: n = K.lattice_dim()
-        sage: L = ToricLattice(n**2)
-        sage: W = VectorSpace(F, n**2)
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: pi_of_K_star = positive_operator_gens(K.dual())
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: pi_star = Cone([p.list() for p in pi_of_K_star],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: M = MatrixSpace(F, n)
-        sage: L = M(pi_cone.random_element(ring=QQ).list())
-        sage: pi_star.contains(W(L.transpose().list()))
-        True
-
-        sage: L = W.random_element()
-        sage: L_star = W(M(L.list()).transpose().list())
-        sage: pi_cone.contains(L) ==  pi_star.contains(L_star)
-        True
-
-    The Lyapunov rank of the positive operator cone is the product of
-    the Lyapunov ranks of the associated cones if they're all proper::
-
-        sage: K1 = random_cone(max_ambient_dim=4,
-        ....:                  strictly_convex=True,
-        ....:                  solid=True)
-        sage: K2 = random_cone(max_ambient_dim=4,
-        ....:                  strictly_convex=True,
-        ....:                  solid=True)
-        sage: pi_K1_K2 = positive_operator_gens(K1,K2)
-        sage: L = ToricLattice(K1.lattice_dim() * K2.lattice_dim())
-        sage: pi_cone = Cone([ g.list() for g in pi_K1_K2 ],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: beta1 = K1.lyapunov_rank()
-        sage: beta2 = K2.lyapunov_rank()
-        sage: pi_cone.lyapunov_rank() == beta1*beta2
-        True
-
-    The Lyapunov-like operators on a proper polyhedral positive operator
-    cone can be computed from the Lyapunov-like operators on the cones
-    with respect to which the operators are positive::
-
-        sage: K1 = random_cone(max_ambient_dim=4,
-        ....:                  strictly_convex=True,
-        ....:                  solid=True)
-        sage: K2 = random_cone(max_ambient_dim=4,
-        ....:                  strictly_convex=True,
-        ....:                  solid=True)
-        sage: pi_K1_K2 = positive_operator_gens(K1,K2)
-        sage: F = K1.lattice().base_field()
-        sage: m = K1.lattice_dim()
-        sage: n = K2.lattice_dim()
-        sage: L = ToricLattice(m*n)
-        sage: M1 = MatrixSpace(F, m, m)
-        sage: M2 = MatrixSpace(F, n, n)
-        sage: LL_K1 = [ M1(x.list()) for x in K1.dual().lyapunov_like_basis() ]
-        sage: LL_K2 = [ M2(x.list()) for x in K2.lyapunov_like_basis() ]
-        sage: tps = [ s.tensor_product(x) for x in LL_K1 for s in LL_K2 ]
-        sage: W = VectorSpace(F, (m**2)*(n**2))
-        sage: expected = span(F, [ W(x.list()) for x in tps ])
-        sage: pi_cone = Cone([p.list() for p in pi_K1_K2],
-        ....:                 lattice=L,
-        ....:                 check=False)
-        sage: LL_pi = pi_cone.lyapunov_like_basis()
-        sage: actual = span(F, [ W(x.list()) for x in LL_pi ])
-        sage: actual == expected
-        True
-
-    """
-    if K2 is None:
-        K2 = K1
-
-    # Matrices are not vectors in Sage, so we have to convert them
-    # to vectors explicitly before we can find a basis. We need these
-    # two values to construct the appropriate "long vector" space.
-    F = K1.lattice().base_field()
-    n = K1.lattice_dim()
-    m = K2.lattice_dim()
-
-    tensor_products = [ s.tensor_product(x) for x in K1 for s in K2.dual() ]
-
-    # Convert those tensor products to long vectors.
-    W = VectorSpace(F, n*m)
-    vectors = [ W(tp.list()) for tp in tensor_products ]
-
-    check = True
-    if K1.is_proper() and K2.is_proper():
-        # All of the generators involved are extreme vectors and
-        # therefore minimal. If this cone is neither solid nor
-        # strictly convex, then the tensor product of ``s`` and ``x``
-        # is the same as that of ``-s`` and ``-x``. However, as a
-        # /set/, ``tensor_products`` may still be minimal.
-        check = False
-
-    # Create the dual cone of the positive operators, expressed as
-    # long vectors.
-    pi_dual = Cone(vectors, ToricLattice(W.dimension()), check=check)
-
-    # Now compute the desired cone from its dual...
-    pi_cone = pi_dual.dual()
-
-    # And finally convert its rays back to matrix representations.
-    M = MatrixSpace(F, m, n)
-    return [ M(v.list()) for v in pi_cone ]
-
-
-def cross_positive_operator_gens(K):
-    r"""
-    Compute generators of the cone of cross-positive operators on this
-    cone.
-
-    OUTPUT:
-
-    A list of `n`-by-``n`` matrices where ``n == K.lattice_dim()``.
-    Each matrix ``L`` in the list should have the property that
-    ``(L*x).inner_product(s) >= 0`` whenever ``(x,s)`` is an element of
-    this cone's :meth:`discrete_complementarity_set`. Moreover, any
-    conic (nonnegative linear) combination of these matrices shares the
-    same property.
-
-    .. SEEALSO::
-
-       :meth:`positive_operator_gens`, :meth:`Z_operator_gens`,
-
-    EXAMPLES:
-
-    Cross-positive operators on the nonnegative orthant are negations
-    of Z-matrices; that is, matrices whose off-diagonal elements are
-    nonnegative::
-
-        sage: K = Cone([(1,0),(0,1)])
-        sage: cross_positive_operator_gens(K)
-        [
-        [0 1]  [0 0]  [1 0]  [-1  0]  [0 0]  [ 0  0]
-        [0 0], [1 0], [0 0], [ 0  0], [0 1], [ 0 -1]
-        ]
-        sage: K = Cone([(1,0,0,0),(0,1,0,0),(0,0,1,0),(0,0,0,1)])
-        sage: all([ c[i][j] >= 0 for c in cross_positive_operator_gens(K)
-        ....:                    for i in range(c.nrows())
-        ....:                    for j in range(c.ncols())
-        ....:                    if i != j ])
-        True
-
-    The trivial cone in a trivial space has no cross-positive operators::
-
-        sage: K = Cone([], ToricLattice(0))
-        sage: cross_positive_operator_gens(K)
-        []
-
-    Every operator is a cross-positive operator on the ambient vector
-    space::
-
-        sage: K = Cone([(1,),(-1,)])
-        sage: K.is_full_space()
-        True
-        sage: cross_positive_operator_gens(K)
-        [[1], [-1]]
-
-        sage: K = Cone([(1,0),(-1,0),(0,1),(0,-1)])
-        sage: K.is_full_space()
-        True
-        sage: cross_positive_operator_gens(K)
-        [
-        [1 0]  [-1  0]  [0 1]  [ 0 -1]  [0 0]  [ 0  0]  [0 0]  [ 0  0]
-        [0 0], [ 0  0], [0 0], [ 0  0], [1 0], [-1  0], [0 1], [ 0 -1]
-        ]
-
-    A non-obvious application is to find the cross-positive operators
-    on the right half-plane::
-
-        sage: K = Cone([(1,0),(0,1),(0,-1)])
-        sage: cross_positive_operator_gens(K)
-        [
-        [1 0]  [-1  0]  [0 0]  [ 0  0]  [0 0]  [ 0  0]
-        [0 0], [ 0  0], [1 0], [-1  0], [0 1], [ 0 -1]
-        ]
-
-    Cross-positive operators on a subspace are Lyapunov-like and
-    vice-versa::
-
-        sage: K = Cone([(1,0),(-1,0),(0,1),(0,-1)])
-        sage: K.is_full_space()
-        True
-        sage: lls = span([ vector(l.list()) for l in K.lyapunov_like_basis() ])
-        sage: cs  = span([ vector(c.list()) for c in cross_positive_operator_gens(K) ])
-        sage: cs == lls
-        True
-
-    TESTS:
-
-    The cross-positive property is possessed by every cross-positive
-    operator::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: dcs = K.discrete_complementarity_set()
-        sage: all([(c*x).inner_product(s) >= 0 for c in Sigma_of_K
-        ....:                                  for (x,s) in dcs])
-        True
-
-    The lineality space of the cone of cross-positive operators is the
-    space of Lyapunov-like operators::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: L = ToricLattice(K.lattice_dim()**2)
-        sage: Sigma_cone = Cone([ c.list() for c in cross_positive_operator_gens(K) ],
-        ....:                     lattice=L,
-        ....:                     check=False)
-        sage: ll_basis = [ vector(l.list()) for l in K.lyapunov_like_basis() ]
-        sage: lls = L.vector_space().span(ll_basis)
-        sage: Sigma_cone.linear_subspace() == lls
-        True
-
-    The lineality of the cross-positive operators on a cone is the
-    Lyapunov rank of that cone::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: L = ToricLattice(K.lattice_dim()**2)
-        sage: Sigma_cone  = Cone([ c.list() for c in Sigma_of_K ],
-        ....:                      lattice=L,
-        ....:                      check=False)
-        sage: Sigma_cone.lineality() == K.lyapunov_rank()
-        True
-
-    The lineality spaces of the duals of the positive and cross-positive
-    operator cones are equal. From this it follows that the dimensions of
-    the cross-positive operator cone and positive operator cone are equal::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: pi_of_K = positive_operator_gens(K)
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: L = ToricLattice(K.lattice_dim()**2)
-        sage: pi_cone = Cone([p.list() for p in pi_of_K],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: Sigma_cone = Cone([ c.list() for c in Sigma_of_K],
-        ....:                     lattice=L,
-        ....:                     check=False)
-        sage: pi_cone.dim() == Sigma_cone.dim()
-        True
-        sage: pi_star = pi_cone.dual()
-        sage: sigma_star = Sigma_cone.dual()
-        sage: pi_star.linear_subspace() == sigma_star.linear_subspace()
-        True
-
-    The trivial cone, full space, and half-plane all give rise to the
-    expected dimensions::
-
-        sage: n = ZZ.random_element().abs()
-        sage: K = Cone([[0] * n], ToricLattice(n))
-        sage: K.is_trivial()
-        True
-        sage: L = ToricLattice(n^2)
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: Sigma_cone = Cone([c.list() for c in Sigma_of_K],
-        ....:                    lattice=L,
-        ....:                    check=False)
-        sage: actual = Sigma_cone.dim()
-        sage: actual == n^2
-        True
-        sage: K = K.dual()
-        sage: K.is_full_space()
-        True
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: Sigma_cone = Cone([ c.list() for c in Sigma_of_K ],
-        ....:                     lattice=L,
-        ....:                     check=False)
-        sage: actual = Sigma_cone.dim()
-        sage: actual == n^2
-        True
-        sage: K = Cone([(1,0),(0,1),(0,-1)])
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: Sigma_cone = Cone([ c.list() for c in Sigma_of_K ], check=False)
-        sage: Sigma_cone.dim() == 3
-        True
-
-    The cross-positive operators of a permuted cone can be obtained by
-    conjugation::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: L = ToricLattice(K.lattice_dim()**2)
-        sage: p = SymmetricGroup(K.lattice_dim()).random_element().matrix()
-        sage: pK = Cone([ p*k for k in K ], K.lattice(), check=False)
-        sage: Sigma_of_pK = cross_positive_operator_gens(pK)
-        sage: actual = Cone([t.list() for t in Sigma_of_pK],
-        ....:                lattice=L,
-        ....:                check=False)
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: expected = Cone([ (p*t*p.inverse()).list() for t in Sigma_of_K ],
-        ....:                   lattice=L,
-        ....:                   check=False)
-        sage: actual.is_equivalent(expected)
-        True
-
-    An operator is cross-positive on a cone if and only if its
-    adjoint is cross-positive on the dual of that cone::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: F = K.lattice().vector_space().base_field()
-        sage: n = K.lattice_dim()
-        sage: L = ToricLattice(n**2)
-        sage: W = VectorSpace(F, n**2)
-        sage: Sigma_of_K = cross_positive_operator_gens(K)
-        sage: Sigma_of_K_star = cross_positive_operator_gens(K.dual())
-        sage: Sigma_cone = Cone([ p.list() for p in Sigma_of_K ],
-        ....:                     lattice=L,
-        ....:                     check=False)
-        sage: Sigma_star = Cone([ p.list() for p in Sigma_of_K_star ],
-        ....:                     lattice=L,
-        ....:                     check=False)
-        sage: M = MatrixSpace(F, n)
-        sage: L = M(Sigma_cone.random_element(ring=QQ).list())
-        sage: Sigma_star.contains(W(L.transpose().list()))
-        True
-
-        sage: L = W.random_element()
-        sage: L_star = W(M(L.list()).transpose().list())
-        sage: Sigma_cone.contains(L) ==  Sigma_star.contains(L_star)
-        True
-    """
-    # Matrices are not vectors in Sage, so we have to convert them
-    # to vectors explicitly before we can find a basis. We need these
-    # two values to construct the appropriate "long vector" space.
-    F = K.lattice().base_field()
-    n = K.lattice_dim()
-
-    # These tensor products contain generators for the dual cone of
-    # the cross-positive operators.
-    tensor_products = [ s.tensor_product(x)
-                        for (x,s) in K.discrete_complementarity_set() ]
-
-    # Turn our matrices into long vectors...
-    W = VectorSpace(F, n**2)
-    vectors = [ W(m.list()) for m in tensor_products ]
-
-    check = True
-    if K.is_proper():
-        # All of the generators involved are extreme vectors and
-        # therefore minimal. If this cone is neither solid nor
-        # strictly convex, then the tensor product of ``s`` and ``x``
-        # is the same as that of ``-s`` and ``-x``. However, as a
-        # /set/, ``tensor_products`` may still be minimal.
-        check = False
-
-    # Create the dual cone of the cross-positive operators,
-    # expressed as long vectors.
-    Sigma_dual = Cone(vectors, lattice=ToricLattice(W.dimension()), check=check)
-
-    # Now compute the desired cone from its dual...
-    Sigma_cone = Sigma_dual.dual()
-
-    # And finally convert its rays back to matrix representations.
-    M = MatrixSpace(F, n)
-    return [ M(v.list()) for v in Sigma_cone ]
-
-
-def Z_operator_gens(K):
-    r"""
-    Compute generators of the cone of Z-operators on this cone.
-
-    The Z-operators on a cone generalize the Z-matrices over the
-    nonnegative orthant. They are simply negations of the
-    :meth:`cross_positive_operators`.
-
-    OUTPUT:
-
-    A list of `n`-by-``n`` matrices where ``n == K.lattice_dim()``.
-    Each matrix ``L`` in the list should have the property that
-    ``(L*x).inner_product(s) <= 0`` whenever ``(x,s)`` is an element of
-    this cone's :meth:`discrete_complementarity_set`. Moreover, any
-    conic (nonnegative linear) combination of these matrices shares the
-    same property.
-
-    .. SEEALSO::
-
-       :meth:`positive_operator_gens`, :meth:`cross_positive_operator_gens`,
-
-    TESTS:
-
-    The Z-property is possessed by every Z-operator::
-
-        sage: set_random_seed()
-        sage: K = random_cone(max_ambient_dim=4)
-        sage: Z_of_K = Z_operator_gens(K)
-        sage: dcs = K.discrete_complementarity_set()
-        sage: all([(z*x).inner_product(s) <= 0 for z in Z_of_K
-        ....:                                  for (x,s) in dcs])
-        True
-    """
-    return [ -cp for cp in cross_positive_operator_gens(K) ]
-
-
 def LL_cone(K):
     gens = K.lyapunov_like_basis()
     L = ToricLattice(K.lattice_dim()**2)
     return Cone([ g.list() for g in gens ], lattice=L, check=False)
 
 def Sigma_cone(K):
-    gens = cross_positive_operator_gens(K)
+    gens = K.cross_positive_operator_gens()
     L = ToricLattice(K.lattice_dim()**2)
     return Cone([ g.list() for g in gens ], lattice=L, check=False)
 
 def Z_cone(K):
-    gens = Z_operator_gens(K)
+    gens = K.Z_operator_gens()
     L = ToricLattice(K.lattice_dim()**2)
     return Cone([ g.list() for g in gens ], lattice=L, check=False)
 
-def pi_cone(K):
-    gens = positive_operator_gens(K)
-    L = ToricLattice(K.lattice_dim()**2)
+def pi_cone(K1, K2=None):
+    if K2 is None:
+        K2 = K1
+    gens = K1.positive_operator_gens(K2)
+    L = ToricLattice(K1.lattice_dim()*K2.lattice_dim())
     return Cone([ g.list() for g in gens ], lattice=L, check=False)