mjo/cone/cone.py

   1 from sage.all import *
   2
   3 def is_lyapunov_like(L,K):
   4     r"""
   5     Determine whether or not ``L`` is Lyapunov-like on ``K``.
   6
   7     We say that ``L`` is Lyapunov-like on ``K`` if `\left\langle
   8     L\left\lparenx\right\rparen,s\right\rangle = 0` for all pairs
   9     `\left\langle x,s \right\rangle` in the complementarity set of
  10     ``K``. It is known [Orlitzky]_ that this property need only be
  11     checked for generators of ``K`` and its dual.
  12
  13     INPUT:
  14
  15     - ``L`` -- A linear transformation or matrix.
  16
  17     - ``K`` -- A polyhedral closed convex cone.
  18
  19     OUTPUT:
  20
  21     ``True`` if it can be proven that ``L`` is Lyapunov-like on ``K``,
  22     and ``False`` otherwise.
  23
  24     .. WARNING::
  25
  26         If this function returns ``True``, then ``L`` is Lyapunov-like
  27         on ``K``. However, if ``False`` is returned, that could mean one
  28         of two things. The first is that ``L`` is definitely not
  29         Lyapunov-like on ``K``. The second is more of an "I don't know"
  30         answer, returned (for example) if we cannot prove that an inner
  31         product is zero.
  32
  33     REFERENCES:
  34
  35     M. Orlitzky. The Lyapunov rank of an improper cone.
  36     http://www.optimization-online.org/DB_HTML/2015/10/5135.html
  37
  38     EXAMPLES:
  39
  40     The identity is always Lyapunov-like in a nontrivial space::
  41
  42         sage: set_random_seed()
  43         sage: K = random_cone(min_ambient_dim=1, max_ambient_dim=8)
  44         sage: L = identity_matrix(K.lattice_dim())
  45         sage: is_lyapunov_like(L,K)
  46         True
  47
  48     As is the "zero" transformation::
  49
  50         sage: K = random_cone(min_ambient_dim=1, max_ambient_dim=8)
  51         sage: R = K.lattice().vector_space().base_ring()
  52         sage: L = zero_matrix(R, K.lattice_dim())
  53         sage: is_lyapunov_like(L,K)
  54         True
  55
  56         Everything in ``K.lyapunov_like_basis()`` should be Lyapunov-like
  57         on ``K``::
  58
  59         sage: K = random_cone(min_ambient_dim=1, max_ambient_dim=6)
  60         sage: all([ is_lyapunov_like(L,K) for L in K.lyapunov_like_basis() ])
  61         True
  62
  63     """
  64     return all([(L*x).inner_product(s) == 0
  65                 for (x,s) in K.discrete_complementarity_set()])
  66
  67
  68 def motzkin_decomposition(K):
  69     r"""
  70     Return the pair of components in the Motzkin decomposition of this cone.
  71
  72     Every convex cone is the direct sum of a strictly convex cone and a
  73     linear subspace [Stoer-Witzgall]_. Return a pair ``(P,S)`` of cones
  74     such that ``P`` is strictly convex, ``S`` is a subspace, and ``K``
  75     is the direct sum of ``P`` and ``S``.
  76
  77     OUTPUT:
  78
  79     An ordered pair ``(P,S)`` of closed convex polyhedral cones where
  80     ``P`` is strictly convex, ``S`` is a subspace, and ``K`` is the
  81     direct sum of ``P`` and ``S``.
  82
  83     REFERENCES:
  84
  85     .. [Stoer-Witzgall] J. Stoer and C. Witzgall. Convexity and
  86        Optimization in Finite Dimensions I. Springer-Verlag, New
  87        York, 1970.
  88
  89     EXAMPLES:
  90
  91     The nonnegative orthant is strictly convex, so it is its own
  92     strictly convex component and its subspace component is trivial::
  93
  94         sage: K = Cone([(1,0,0),(0,1,0),(0,0,1)])
  95         sage: (P,S) = motzkin_decomposition(K)
  96         sage: K.is_equivalent(P)
  97         True
  98         sage: S.is_trivial()
  99         True
 100
 101     Likewise, full spaces are their own subspace components::
 102
 103         sage: K = Cone([(1,0),(-1,0),(0,1),(0,-1)])
 104         sage: K.is_full_space()
 105         True
 106         sage: (P,S) = motzkin_decomposition(K)
 107         sage: K.is_equivalent(S)
 108         True
 109         sage: P.is_trivial()
 110         True
 111
 112     TESTS:
 113
 114     A random point in the cone should belong to either the strictly
 115     convex component or the subspace component. If the point is nonzero,
 116     it cannot be in both::
 117
 118         sage: set_random_seed()
 119         sage: K = random_cone(max_ambient_dim=8)
 120         sage: (P,S) = motzkin_decomposition(K)
 121         sage: x = K.random_element()
 122         sage: P.contains(x) or S.contains(x)
 123         True
 124         sage: x.is_zero() or (P.contains(x) != S.contains(x))
 125         True
 126
 127     The strictly convex component should always be strictly convex, and
 128     the subspace component should always be a subspace::
 129
 130         sage: set_random_seed()
 131         sage: K = random_cone(max_ambient_dim=8)
 132         sage: (P,S) = motzkin_decomposition(K)
 133         sage: P.is_strictly_convex()
 134         True
 135         sage: S.lineality() == S.dim()
 136         True
 137
 138     The generators of the components are obtained from orthogonal
 139     projections of the original generators [Stoer-Witzgall]_::
 140
 141         sage: set_random_seed()
 142         sage: K = random_cone(max_ambient_dim=8)
 143         sage: (P,S) = motzkin_decomposition(K)
 144         sage: A = S.linear_subspace().complement().matrix()
 145         sage: proj_S_perp = A.transpose() * (A*A.transpose()).inverse() * A
 146         sage: expected_P = Cone([ proj_S_perp*g for g in K ], K.lattice())
 147         sage: P.is_equivalent(expected_P)
 148         True
 149         sage: A = S.linear_subspace().matrix()
 150         sage: proj_S = A.transpose() * (A*A.transpose()).inverse() * A
 151         sage: expected_S = Cone([ proj_S*g for g in K ], K.lattice())
 152         sage: S.is_equivalent(expected_S)
 153         True
 154     """
 155     # The lines() method only gives us one generator for each line,
 156     # so we negate the result and combine everything for the full set.
 157     S = Cone([p*l for p in [1,-1] for l in K.lines()], K.lattice())
 158
 159     # Since ``S`` is a subspace, the rays of its dual generate its
 160     # orthogonal complement.
 161     P = K.intersection( Cone(S.dual(), K.lattice()) )
 162
 163     return (P,S)
 164
 165
 166 def positive_operator_gens(K):
 167     r"""
 168     Compute generators of the cone of positive operators on this cone.
 169
 170     OUTPUT:
 171
 172     A list of `n`-by-``n`` matrices where ``n == K.lattice_dim()``.
 173     Each matrix ``P`` in the list should have the property that ``P*x``
 174     is an element of ``K`` whenever ``x`` is an element of
 175     ``K``. Moreover, any nonnegative linear combination of these
 176     matrices shares the same property.
 177
 178     EXAMPLES:
 179
 180     The trivial cone in a trivial space has no positive operators::
 181
 182         sage: K = Cone([], ToricLattice(0))
 183         sage: positive_operator_gens(K)
 184         []
 185
 186     Positive operators on the nonnegative orthant are nonnegative matrices::
 187
 188         sage: K = Cone([(1,)])
 189         sage: positive_operator_gens(K)
 190         [[1]]
 191
 192         sage: K = Cone([(1,0),(0,1)])
 193         sage: positive_operator_gens(K)
 194         [
 195         [1 0]  [0 1]  [0 0]  [0 0]
 196         [0 0], [0 0], [1 0], [0 1]
 197         ]
 198
 199     Every operator is positive on the ambient vector space::
 200
 201         sage: K = Cone([(1,),(-1,)])
 202         sage: K.is_full_space()
 203         True
 204         sage: positive_operator_gens(K)
 205         [[1], [-1]]
 206
 207         sage: K = Cone([(1,0),(-1,0),(0,1),(0,-1)])
 208         sage: K.is_full_space()
 209         True
 210         sage: positive_operator_gens(K)
 211         [
 212         [1 0]  [-1  0]  [0 1]  [ 0 -1]  [0 0]  [ 0  0]  [0 0]  [ 0  0]
 213         [0 0], [ 0  0], [0 0], [ 0  0], [1 0], [-1  0], [0 1], [ 0 -1]
 214         ]
 215
 216     TESTS:
 217
 218     A positive operator on a cone should send its generators into the cone::
 219
 220         sage: set_random_seed()
 221         sage: K = random_cone(max_ambient_dim=5)
 222         sage: pi_of_K = positive_operator_gens(K)
 223         sage: all([K.contains(p*x) for p in pi_of_K for x in K.rays()])
 224         True
 225
 226     The dimension of the cone of positive operators is given by the
 227     corollary in my paper::
 228
 229         sage: set_random_seed()
 230         sage: K = random_cone(max_ambient_dim=5)
 231         sage: n = K.lattice_dim()
 232         sage: m = K.dim()
 233         sage: l = K.lineality()
 234         sage: pi_of_K = positive_operator_gens(K)
 235         sage: L = ToricLattice(n**2)
 236         sage: actual = Cone([p.list() for p in pi_of_K], lattice=L).dim()
 237         sage: expected = n**2 - l*(m - l) - (n - m)*m
 238         sage: actual == expected
 239         True
 240
 241     The lineality of the cone of positive operators is given by the
 242     corollary in my paper::
 243
 244         sage: set_random_seed()
 245         sage: K = random_cone(max_ambient_dim=5)
 246         sage: n = K.lattice_dim()
 247         sage: pi_of_K = positive_operator_gens(K)
 248         sage: L = ToricLattice(n**2)
 249         sage: actual = Cone([p.list() for p in pi_of_K], lattice=L).lineality()
 250         sage: expected = n**2 - K.dim()*K.dual().dim()
 251         sage: actual == expected
 252         True
 253
 254     The cone ``K`` is proper if and only if the cone of positive
 255     operators on ``K`` is proper::
 256
 257         sage: set_random_seed()
 258         sage: K = random_cone(max_ambient_dim=5)
 259         sage: pi_of_K = positive_operator_gens(K)
 260         sage: L = ToricLattice(K.lattice_dim()**2)
 261         sage: pi_cone = Cone([p.list() for p in pi_of_K], lattice=L)
 262         sage: K.is_proper() == pi_cone.is_proper()
 263         True
 264     """
 265     # Matrices are not vectors in Sage, so we have to convert them
 266     # to vectors explicitly before we can find a basis. We need these
 267     # two values to construct the appropriate "long vector" space.
 268     F = K.lattice().base_field()
 269     n = K.lattice_dim()
 270
 271     tensor_products = [ s.tensor_product(x) for x in K for s in K.dual() ]
 272
 273     # Convert those tensor products to long vectors.
 274     W = VectorSpace(F, n**2)
 275     vectors = [ W(tp.list()) for tp in tensor_products ]
 276
 277     # Create the *dual* cone of the positive operators, expressed as
 278     # long vectors..
 279     pi_dual = Cone(vectors, ToricLattice(W.dimension()))
 280
 281     # Now compute the desired cone from its dual...
 282     pi_cone = pi_dual.dual()
 283
 284     # And finally convert its rays back to matrix representations.
 285     M = MatrixSpace(F, n)
 286     return [ M(v.list()) for v in pi_cone.rays() ]
 287
 288
 289 def Z_transformation_gens(K):
 290     r"""
 291     Compute generators of the cone of Z-transformations on this cone.
 292
 293     OUTPUT:
 294
 295     A list of `n`-by-``n`` matrices where ``n == K.lattice_dim()``.
 296     Each matrix ``L`` in the list should have the property that
 297     ``(L*x).inner_product(s) <= 0`` whenever ``(x,s)`` is an element the
 298     discrete complementarity set of ``K``. Moreover, any nonnegative
 299     linear combination of these matrices shares the same property.
 300
 301     EXAMPLES:
 302
 303     Z-transformations on the nonnegative orthant are just Z-matrices.
 304     That is, matrices whose off-diagonal elements are nonnegative::
 305
 306         sage: K = Cone([(1,0),(0,1)])
 307         sage: Z_transformation_gens(K)
 308         [
 309         [ 0 -1]  [ 0  0]  [-1  0]  [1 0]  [ 0  0]  [0 0]
 310         [ 0  0], [-1  0], [ 0  0], [0 0], [ 0 -1], [0 1]
 311         ]
 312         sage: K = Cone([(1,0,0,0),(0,1,0,0),(0,0,1,0),(0,0,0,1)])
 313         sage: all([ z[i][j] <= 0 for z in Z_transformation_gens(K)
 314         ....:                    for i in range(z.nrows())
 315         ....:                    for j in range(z.ncols())
 316         ....:                    if i != j ])
 317         True
 318
 319     The trivial cone in a trivial space has no Z-transformations::
 320
 321         sage: K = Cone([], ToricLattice(0))
 322         sage: Z_transformation_gens(K)
 323         []
 324
 325     Z-transformations on a subspace are Lyapunov-like and vice-versa::
 326
 327         sage: K = Cone([(1,0),(-1,0),(0,1),(0,-1)])
 328         sage: K.is_full_space()
 329         True
 330         sage: lls = span([ vector(l.list()) for l in K.lyapunov_like_basis() ])
 331         sage: zs  = span([ vector(z.list()) for z in Z_transformation_gens(K) ])
 332         sage: zs == lls
 333         True
 334
 335     TESTS:
 336
 337     The Z-property is possessed by every Z-transformation::
 338
 339         sage: set_random_seed()
 340         sage: K = random_cone(max_ambient_dim=6)
 341         sage: Z_of_K = Z_transformation_gens(K)
 342         sage: dcs = K.discrete_complementarity_set()
 343         sage: all([(z*x).inner_product(s) <= 0 for z in Z_of_K
 344         ....:                                  for (x,s) in dcs])
 345         True
 346
 347     The lineality space of Z is LL::
 348
 349         sage: set_random_seed()
 350         sage: K = random_cone(min_ambient_dim=1, max_ambient_dim=6)
 351         sage: lls = span([ vector(l.list()) for l in K.lyapunov_like_basis() ])
 352         sage: z_cone  = Cone([ z.list() for z in Z_transformation_gens(K) ])
 353         sage: z_cone.linear_subspace() == lls
 354         True
 355
 356     And thus, the lineality of Z is the Lyapunov rank::
 357
 358         sage: set_random_seed()
 359         sage: K = random_cone(max_ambient_dim=6)
 360         sage: Z_of_K = Z_transformation_gens(K)
 361         sage: L = ToricLattice(K.lattice_dim()**2)
 362         sage: z_cone  = Cone([ z.list() for z in Z_of_K ], lattice=L)
 363         sage: z_cone.lineality() == K.lyapunov_rank()
 364         True
 365
 366     The lineality spaces of pi-star and Z-star are equal:
 367
 368         sage: set_random_seed()
 369         sage: K = random_cone(max_ambient_dim=5)
 370         sage: pi_of_K = positive_operator_gens(K)
 371         sage: Z_of_K = Z_transformation_gens(K)
 372         sage: L = ToricLattice(K.lattice_dim()**2)
 373         sage: pi_star = Cone([p.list() for p in pi_of_K], lattice=L).dual()
 374         sage: z_star  = Cone([ z.list() for z in Z_of_K], lattice=L).dual()
 375         sage: pi_star.linear_subspace() == z_star.linear_subspace()
 376         True
 377     """
 378     # Matrices are not vectors in Sage, so we have to convert them
 379     # to vectors explicitly before we can find a basis. We need these
 380     # two values to construct the appropriate "long vector" space.
 381     F = K.lattice().base_field()
 382     n = K.lattice_dim()
 383
 384     # These tensor products contain generators for the dual cone of
 385     # the cross-positive transformations.
 386     tensor_products = [ s.tensor_product(x)
 387                         for (x,s) in K.discrete_complementarity_set() ]
 388
 389     # Turn our matrices into long vectors...
 390     W = VectorSpace(F, n**2)
 391     vectors = [ W(m.list()) for m in tensor_products ]
 392
 393     # Create the *dual* cone of the cross-positive operators,
 394     # expressed as long vectors..
 395     Sigma_dual = Cone(vectors, lattice=ToricLattice(W.dimension()))
 396
 397     # Now compute the desired cone from its dual...
 398     Sigma_cone = Sigma_dual.dual()
 399
 400     # And finally convert its rays back to matrix representations.
 401     # But first, make them negative, so we get Z-transformations and
 402     # not cross-positive ones.
 403     M = MatrixSpace(F, n)
 404     return [ -M(v.list()) for v in Sigma_cone.rays() ]
 405
 406
 407 def Z_cone(K):
 408     gens = Z_transformation_gens(K)
 409     L = None
 410     if len(gens) == 0:
 411         L = ToricLattice(0)
 412     return Cone([ g.list() for g in gens ], lattice=L)
 413
 414 def pi_cone(K):
 415     gens = positive_operator_gens(K)
 416     L = None
 417     if len(gens) == 0:
 418         L = ToricLattice(0)
 419     return Cone([ g.list() for g in gens ], lattice=L)