Add a general sparse iterative solver: CGNR
This adds a new LinearOperator which implements symmetric
products of a matrix, and a new CGNR solver to leverage
CG to directly solve the normal equations. This also
includes a block diagonal preconditioner. In experiments
on problem-16, the non-preconditioned version is about
1/5 the speed of SPARSE_SCHUR, and the preconditioned
version using block cholesky is about 20% slower than
SPARSE_SCHUR.
diff --git a/internal/ceres/CMakeLists.txt b/internal/ceres/CMakeLists.txt
index 7891138..03d650a 100644
--- a/internal/ceres/CMakeLists.txt
+++ b/internal/ceres/CMakeLists.txt
@@ -29,6 +29,7 @@
# Author: keir@google.com (Keir Mierle)
SET(CERES_INTERNAL_SRC
+ block_diagonal_preconditioner.cc
block_evaluate_preparer.cc
block_jacobian_writer.cc
block_random_access_dense_matrix.cc
@@ -37,6 +38,7 @@
block_sparse_matrix.cc
block_structure.cc
canonical_views_clustering.cc
+ cgnr_solver.cc
compressed_row_jacobian_writer.cc
compressed_row_sparse_matrix.cc
conditioned_cost_function.cc
