github.com/gopherd/gonum@v0.0.4/lapack/gonum/dpbtf2.go (about)

     1  // Copyright ©2017 The Gonum Authors. All rights reserved.
     2  // Use of this source code is governed by a BSD-style
     3  // license that can be found in the LICENSE file.
     4  
     5  package gonum
     6  
     7  import (
     8  	"math"
     9  
    10  	"github.com/gopherd/gonum/blas"
    11  	"github.com/gopherd/gonum/blas/blas64"
    12  )
    13  
    14  // Dpbtf2 computes the Cholesky factorization of a symmetric positive banded
    15  // matrix ab. The matrix ab is n×n with kd diagonal bands. The Cholesky
    16  // factorization computed is
    17  //  A = Uᵀ * U  if ul == blas.Upper
    18  //  A = L * Lᵀ  if ul == blas.Lower
    19  // ul also specifies the storage of ab. If ul == blas.Upper, then
    20  // ab is stored as an upper-triangular banded matrix with kd super-diagonals,
    21  // and if ul == blas.Lower, ab is stored as a lower-triangular banded matrix
    22  // with kd sub-diagonals. On exit, the banded matrix U or L is stored in-place
    23  // into ab depending on the value of ul. Dpbtf2 returns whether the factorization
    24  // was successfully completed.
    25  //
    26  // The band storage scheme is illustrated below when n = 6, and kd = 2.
    27  // The resulting Cholesky decomposition is stored in the same elements as the
    28  // input band matrix (a11 becomes u11 or l11, etc.).
    29  //
    30  //  ul = blas.Upper
    31  //  a11 a12 a13
    32  //  a22 a23 a24
    33  //  a33 a34 a35
    34  //  a44 a45 a46
    35  //  a55 a56  *
    36  //  a66  *   *
    37  //
    38  //  ul = blas.Lower
    39  //   *   *  a11
    40  //   *  a21 a22
    41  //  a31 a32 a33
    42  //  a42 a43 a44
    43  //  a53 a54 a55
    44  //  a64 a65 a66
    45  //
    46  // Dpbtf2 is the unblocked version of the algorithm, see Dpbtrf for the blocked
    47  // version.
    48  //
    49  // Dpbtf2 is an internal routine, exported for testing purposes.
    50  func (Implementation) Dpbtf2(uplo blas.Uplo, n, kd int, ab []float64, ldab int) (ok bool) {
    51  	switch {
    52  	case uplo != blas.Upper && uplo != blas.Lower:
    53  		panic(badUplo)
    54  	case n < 0:
    55  		panic(nLT0)
    56  	case kd < 0:
    57  		panic(kdLT0)
    58  	case ldab < kd+1:
    59  		panic(badLdA)
    60  	}
    61  
    62  	// Quick return if possible.
    63  	if n == 0 {
    64  		return true
    65  	}
    66  
    67  	if len(ab) < (n-1)*ldab+kd+1 {
    68  		panic(shortAB)
    69  	}
    70  
    71  	bi := blas64.Implementation()
    72  
    73  	kld := max(1, ldab-1)
    74  	if uplo == blas.Upper {
    75  		// Compute the Cholesky factorization A = Uᵀ * U.
    76  		for j := 0; j < n; j++ {
    77  			// Compute U(j,j) and test for non-positive-definiteness.
    78  			ajj := ab[j*ldab]
    79  			if ajj <= 0 {
    80  				return false
    81  			}
    82  			ajj = math.Sqrt(ajj)
    83  			ab[j*ldab] = ajj
    84  			// Compute elements j+1:j+kn of row j and update the trailing submatrix
    85  			// within the band.
    86  			kn := min(kd, n-j-1)
    87  			if kn > 0 {
    88  				bi.Dscal(kn, 1/ajj, ab[j*ldab+1:], 1)
    89  				bi.Dsyr(blas.Upper, kn, -1, ab[j*ldab+1:], 1, ab[(j+1)*ldab:], kld)
    90  			}
    91  		}
    92  		return true
    93  	}
    94  	// Compute the Cholesky factorization A = L * Lᵀ.
    95  	for j := 0; j < n; j++ {
    96  		// Compute L(j,j) and test for non-positive-definiteness.
    97  		ajj := ab[j*ldab+kd]
    98  		if ajj <= 0 {
    99  			return false
   100  		}
   101  		ajj = math.Sqrt(ajj)
   102  		ab[j*ldab+kd] = ajj
   103  		// Compute elements j+1:j+kn of column j and update the trailing submatrix
   104  		// within the band.
   105  		kn := min(kd, n-j-1)
   106  		if kn > 0 {
   107  			bi.Dscal(kn, 1/ajj, ab[(j+1)*ldab+kd-1:], kld)
   108  			bi.Dsyr(blas.Lower, kn, -1, ab[(j+1)*ldab+kd-1:], kld, ab[(j+1)*ldab+kd:], kld)
   109  		}
   110  	}
   111  	return true
   112  }