ceres之cholesky求解器

normal_sparse_cholesky:

dense_cholesky:

normal_sparse_cholesky:

輸入：a,b,d,求解（a'*a+d'*d）x = a'*b

主調函式：

dynamicsparsenormalcholeskysolver::dynamicsparsenormalcholeskysolver(

const linearsolver::options& options)
: options_(options) {}
linearsolver::summary dynamicsparsenormalcholeskysolver::solveimpl(
compressedrowsparsematrix* a,
const double* b,
const linearsolver::persolveoptions& per_solve_options,
double* x)  else 
}  linearsolver::summary summary;
switch (options_.sparse_linear_algebra_library_type) 
if (per_solve_options.d != null) 
return summary;
}

a->leftmultiply(b, x);

2）若d不等於0:a = a+diag(d)

3)solve(ax=b):

以eigen求解為例：

linearsolver::summary dynamicsparsenormalcholeskysolver::solveimplusingeigen(

compressedrowsparsematrix* a, double* rhs_and_solution) 
event_logger.addevent("analyze");
if (solver.info() != eigen::success) 
solver.factorize(lhs);
event_logger.addevent("factorize");
if (solver.info() != eigen::success) 
const vector rhs = vectorref(rhs_and_solution, lhs.cols());
vectorref(rhs_and_solution, lhs.cols()) = solver.solve(rhs);
event_logger.addevent("solve");
if (solver.info() != eigen::success) 
return summary;
#endif  // ceres_use_eigen_sparse
}

(3.2)通過eigen庫內置的cholesky求解器：

eigen::simplicialldlt> solver

求解ax=b得到x值

dense_cholesky：

linearsolver::summary densenormalcholeskysolver::solveimpl(

densesparsematrix* a,
const double* b,
const linearsolver::persolveoptions& per_solve_options,
double* x)  else 
}linearsolver::summary densenormalcholeskysolver::solveusingeigen(
densesparsematrix* a,
const double* b,
const linearsolver::persolveoptions& per_solve_options,
double* x) 
event_logger.addevent("product");
linearsolver::summary summary;
summary.num_iterations = 1;
summary.termination_type = linear_solver_success;
eigen::lltllt =
lhs.selfadjointview().llt();
if (llt.info() != eigen::success)  else 
vectorref(x, num_cols) = llt.solve(rhs);
event_logger.addevent("solve");
return summary;
}

eigen::lltllt =lhs.selfadjointview().llt();

Ceres學習筆記之CMakeLists寫法總結

高博說 不要長期徘徊在自己的舒適區里猶豫不決，這樣是沒有進步的。這句話開啟了我的slam後端優化學習之路。本文從cmakelists開始，總結常用的各個庫在cmakelists中的寫法。眾人 哇！又是這麼小兒科的嗎？博主 哎呀 不要這樣講嘛 qwq我會一直努力的 通常的cmakelists結構是 c...

通過前代法後代法列主元Cholesky求解Ax b

矩陣分解 function a matrix builder n 對於乙個n，我們按課本120頁要求生成 n 1 2 n 1 2的矩陣 a diag repmat 4 1,n 1 2 diag repmat 1 1,n n 2 1 diag repmat 1 1,n n 2 1 diag repma...

遞迴之求冪

剛開始學習用遞迴實現x的n次方時，其思想 或者遞推式 一般如下 x n x x n 1 n 0 x n 1 n 0 c語言 如下 int64為有符號8位元組整數 int64 power int x,int n return x power x,n 1 這種簡單的求冪演算法的時間複雜度為o n 下面介...