Adding the possiblity to use a Cholesky decomposition of the normal equation

during the computation of the Hessian.

sources/plugins/nonlinear_fitter_ceres/fitter.cpp

@@ -132,7 +132,16 @@ bool nonlinear_fitter_ceres::fit_data(const data* d, function* fit, const argume
 	 {
 		 options.max_num_iterations = args.get_int("ceres-max-num-iterations", 50); // Default value = 50
 	 }
-	 options.linear_solver_type = ceres::DENSE_QR;
+	 
+	 if(args["ceres-factorization"] == "normal-cholesky")
+	 {
+		 options.linear_solver_type = ceres::DENSE_NORMAL_CHOLESKY;
+	 }
+	 else
+	 {
+		 options.linear_solver_type = ceres::DENSE_QR;
+	 }
+
 	 if(args.is_defined("ceres-debug"))
 	 {
 		 options.minimizer_progress_to_stdout = true; // Default value = false;

sources/plugins/nonlinear_fitter_ceres/fitter.h

@@ -45,6 +45,8 @@
  *  <ul>
  *		<li><b>--ceres-max-num-iterations</b> <em>[int]</em> to control the number
  *		of iterations the non linear solver will take before returning a solution</li>
+ *		<li><b>--ceres-factorizer</b> <em>[string]</em> to control the type of dense
+ *		factorization method used to solve the <a href="http://homes.cs.washington.edu/~sagarwal/ceres-solver/stable/solving.html?highlight=dense_qr#equation-normal">normal equations</a></li>
  *  </ul>
  */
 class nonlinear_fitter_ceres: public fitter