xref: /petsc/src/mat/impls/centering/centering.c (revision 8bb6a241cc2b8f6ed3ced698b43bc00095eff8a0)
1 #include <petsc/private/matimpl.h> /*I "petscmat.h" I*/
2 
MatMult_Centering(Mat A,Vec xx,Vec yy)3 static PetscErrorCode MatMult_Centering(Mat A, Vec xx, Vec yy)
4 {
5   PetscScalar       *y;
6   const PetscScalar *x;
7   PetscScalar        sum, mean;
8   PetscInt           i, m = A->rmap->n, size;
9 
10   PetscFunctionBegin;
11   PetscCall(VecSum(xx, &sum));
12   PetscCall(VecGetSize(xx, &size));
13   mean = sum / (PetscScalar)size;
14   PetscCall(VecGetArrayRead(xx, &x));
15   PetscCall(VecGetArray(yy, &y));
16   for (i = 0; i < m; i++) y[i] = x[i] - mean;
17   PetscCall(VecRestoreArrayRead(xx, &x));
18   PetscCall(VecRestoreArray(yy, &y));
19   PetscFunctionReturn(PETSC_SUCCESS);
20 }
21 
22 /*MC
23   MATCENTERING - matrix object that implements the (symmetric and idempotent) centering matrix, $ I - (1/N) * 1^T * 1$
24 
25   Level: beginner
26 
27   Note:
28   See `MatCreateCentering()`
29 
30 .seealso:  `Mat`, `MatCreateCentering()`
31 M*/
32 
33 /*@
34   MatCreateCentering - Creates a new matrix object that implements the (symmetric and idempotent) centering matrix, $ I - (1/N) * 1^T * 1$
35 
36   Collective
37 
38   Input Parameters:
39 + comm - MPI communicator
40 . n    - number of local rows (or `PETSC_DECIDE` to have calculated if `N` is given)
41          This value should be the same as the local size used in creating the
42          `y` vector for the matrix-vector product $y = Ax$.
43 - N    - number of global rows (or `PETSC_DETERMINE` to have calculated if `n` is given)
44 
45   Output Parameter:
46 . C - the matrix
47 
48   Notes:
49   The entries of the matrix `C` are not explicitly stored. Instead, the new matrix
50   object is a shell that simply performs the action of the centering matrix, i.e.,
51   multiplying $ C*x$ subtracts the mean of the vector `x` from each of its elements.
52   This is useful for preserving sparsity when mean-centering the columns of a
53   matrix is required. For instance, to perform principal components analysis with
54   a matrix `A`, the composite matrix $C*A$ can be passed to a partial SVD solver.
55 
56   Level: intermediate
57 
58 .seealso: [](ch_matrices), `Mat`, `MatCreateLRC()`, `MatCreateComposite()`
59 @*/
MatCreateCentering(MPI_Comm comm,PetscInt n,PetscInt N,Mat * C)60 PetscErrorCode MatCreateCentering(MPI_Comm comm, PetscInt n, PetscInt N, Mat *C)
61 {
62   PetscMPIInt size;
63 
64   PetscFunctionBegin;
65   PetscCall(MatCreate(comm, C));
66   PetscCall(MatSetSizes(*C, n, n, N, N));
67   PetscCallMPI(MPI_Comm_size(comm, &size));
68   PetscCall(PetscObjectChangeTypeName((PetscObject)*C, MATCENTERING));
69 
70   (*C)->ops->mult                   = MatMult_Centering;
71   (*C)->assembled                   = PETSC_TRUE;
72   (*C)->preallocated                = PETSC_TRUE;
73   (*C)->symmetric                   = PETSC_BOOL3_TRUE;
74   (*C)->symmetry_eternal            = PETSC_TRUE;
75   (*C)->structural_symmetry_eternal = PETSC_TRUE;
76   (*C)->structurally_symmetric      = PETSC_BOOL3_TRUE;
77   if (!PetscDefined(USE_COMPLEX)) (*C)->hermitian = PETSC_BOOL3_TRUE;
78   PetscCall(MatSetUp(*C));
79   PetscFunctionReturn(PETSC_SUCCESS);
80 }
81