|
DenseLinAlgPack: Concreate C++ Classes for Dense Blas-Compatible Linear Algebra
Version of the Day
|
00001 // @HEADER 00002 // *********************************************************************** 00003 // 00004 // Moocho: Multi-functional Object-Oriented arCHitecture for Optimization 00005 // Copyright (2003) Sandia Corporation 00006 // 00007 // Under terms of Contract DE-AC04-94AL85000, there is a non-exclusive 00008 // license for use of this work by or on behalf of the U.S. Government. 00009 // 00010 // Redistribution and use in source and binary forms, with or without 00011 // modification, are permitted provided that the following conditions are 00012 // met: 00013 // 00014 // 1. Redistributions of source code must retain the above copyright 00015 // notice, this list of conditions and the following disclaimer. 00016 // 00017 // 2. Redistributions in binary form must reproduce the above copyright 00018 // notice, this list of conditions and the following disclaimer in the 00019 // documentation and/or other materials provided with the distribution. 00020 // 00021 // 3. Neither the name of the Corporation nor the names of the 00022 // contributors may be used to endorse or promote products derived from 00023 // this software without specific prior written permission. 00024 // 00025 // THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY 00026 // EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 00027 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 00028 // PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE 00029 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 00030 // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 00031 // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 00032 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 00033 // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 00034 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 00035 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 00036 // 00037 // Questions? Contact Roscoe A. Bartlett (rabartl@sandia.gov) 00038 // 00039 // *********************************************************************** 00040 // @HEADER 00041 00042 #include <math.h> 00043 00044 #include "DenseLinAlgPack_InvCholUpdate.hpp" 00045 #include "DenseLinAlgPack_DMatrixClass.hpp" 00046 #include "DenseLinAlgPack_DVectorOp.hpp" 00047 00048 namespace { 00049 // sign function 00050 inline int sign(double v) { 00051 if(v > 0.0) 00052 return 1; 00053 else if(v == 0.0) 00054 return 0; 00055 else 00056 return -1; 00057 } 00058 00059 } 00060 00061 void DenseLinAlgPack::update_chol_factor(DMatrixSlice* pM, DVectorSlice* pu 00062 , const DVectorSlice& v) 00063 { 00064 DMatrixSlice &M = *pM; 00065 DVectorSlice &u = *pu; 00066 00067 assert_gms_square(M); 00068 if(M.rows() != u.dim() || u.dim() != v.dim()) 00069 throw std::length_error("update_chol_factor(): Sizes of matrix and vectors do not match"); 00070 00071 // Algorithm A3.4.1a in Dennis and Schabel 00072 00073 // 0. Set lower diagonal to zero 00074 M.diag(-1) = 0.0; 00075 00076 // 1,2 Find the largest k such that u(k) != 0.0 00077 size_type n = M.rows(), k = n; 00078 { 00079 DVectorSlice::reverse_iterator r_itr_u = u.rbegin(); 00080 while(*r_itr_u == 0.0 && k > 1) --k; 00081 } 00082 00083 // 3. 00084 { 00085 DVectorSlice::reverse_iterator 00086 r_itr_u_i = u(1,k-1).rbegin(), // iterator for u(i), i = k-1,...,1 00087 r_itr_u_ip1 = u(2,k).rbegin(); // iterator for u(i), i = k,...,2 00088 for(size_type i = k-1; i > 0 ; ++r_itr_u_i, ++r_itr_u_ip1, --i) { 00089 value_type u_i = *r_itr_u_i, u_ip1 = *r_itr_u_ip1; 00090 jacobi_rotate(&M, i, u_i, - u_ip1); // 3.1 00091 *r_itr_u_i = (u_i == 0.0) ? ::fabs(u_ip1) : ::sqrt(u_i * u_i + u_ip1 * u_ip1); // 3.2 00092 } 00093 } 00094 00095 // 4. M.row(1) += u(1) * v 00096 DenseLinAlgPack::Vp_StV(&M.row(1), u(1), v); 00097 00098 // 5. 00099 { 00100 DVectorSlice::const_iterator 00101 itr_M_i_i = M.diag().begin(), // iterator for M(i,i), for i = 1,...,k-1 00102 itr_M_ip1_i = M.diag(-1).begin(); // iterator for M(i+1,i), for i = 1,...,k-1 00103 for(size_type i = 1; i < k; ++i) 00104 jacobi_rotate(&M, i, *itr_M_i_i++, - *itr_M_ip1_i++); 00105 } 00106 00107 /* This code does not work 00108 00109 size_type n = M.rows(); 00110 00111 // 0. 00112 { 00113 for(size_type i = 2; i <= n; ++i) 00114 M(i,i-1) = 0; 00115 } 00116 00117 // 1. 00118 size_type k = n; 00119 00120 // 2. 00121 while(u(k) == 0.0 && k > 1) --k; 00122 00123 // 3. 00124 { 00125 for(size_type i = k-1; i >= 1; --i) { 00126 jacobi_rotate(M, i, u(i), - u(i+1)); 00127 if(u(i) == 0) 00128 u(i) = ::fabs(u(i+1)); 00129 else 00130 u(i) = ::sqrt(u(i) * u(i) + u(i+1) * u(i+1)); 00131 } 00132 } 00133 00134 // 4. 00135 { 00136 for(size_type j = 1; j <= n; ++j) 00137 M(1,j) = M(1,j) + dot(u(1),v(j)); 00138 } 00139 00140 // 5. 00141 { 00142 for(size_type i = 1; i <= k-1; ++i) 00143 jacobi_rotate(M, i, M(i,i), - M(i+1,i)); 00144 } 00145 00146 00147 00148 */ 00149 00150 00151 } 00152 00153 void DenseLinAlgPack::jacobi_rotate(DMatrixSlice* pM, size_type row_i, value_type alpha 00154 , value_type beta) 00155 { 00156 DMatrixSlice &M = *pM; 00157 00158 assert_gms_square(M); 00159 00160 // Algorithm A3.4.1a in Dennis and Schabel 00161 00162 // 1. 00163 value_type c, s; 00164 if(alpha == 0.0) { 00165 c = 0.0; 00166 s = sign(beta); 00167 } 00168 else { 00169 value_type den = ::sqrt(alpha * alpha + beta * beta); 00170 c = alpha / den; 00171 s = beta / den; 00172 } 00173 00174 // 2. 00175 size_type i = row_i, n = M.rows(); 00176 00177 // Use iterators instead of element access 00178 DVectorSlice::iterator 00179 itr_M_i = M.row(i)(i,n).begin(), // iterator for M(i,j), for j = i,...,n 00180 itr_M_i_end = M.row(i)(i,n).end(), 00181 itr_M_ip1 = M.row(i+1)(i,n).begin(); // iterator for M(i+1,j), for j = i,...,n 00182 00183 while(itr_M_i != itr_M_i_end) { 00184 value_type y = *itr_M_i, w = *itr_M_ip1; 00185 *itr_M_i++ = c * y - s * w; 00186 *itr_M_ip1++ = s * y + c * w; 00187 } 00188 00189 /* This code does not work 00190 00191 size_type n = M.rows(), i = row_i; 00192 00193 // 1. 00194 value_type c, s; 00195 if(alpha == 0) { 00196 c = 0.0; 00197 s = ::fabs(beta); 00198 } 00199 else { 00200 value_type den = ::sqrt(alpha*alpha + beta*beta); 00201 c = alpha / den; 00202 s = beta / den; 00203 } 00204 00205 // 2. 00206 { 00207 for(size_type j = i; j <= n; ++j) { 00208 size_type y = M(i,j), w = M(i+1,j); 00209 M(i,j) = c*y - s*w; 00210 M(i+1,j) = s*y + c*w; 00211 } 00212 } 00213 00214 00215 00216 */ 00217 00218 }
1.7.6.1