Is Ceres-Solver being used in the factor graph optimization?

I noticed that the output of the optimize function https://www.mathworks.com/help/nav/ref/factorgraph.optimize.html is very similar to that of google's ceres. Is he the backbone for this function?

Answer 1

[Jasvin]

MathWorks' Factor Graph Toolbox is a proprietary software package, and the implementation details of its optimize function are not publicly disclosed. Therefore, it is not possible to say for certain whether Google's Ceres Solver is used as a backbone for this function.

However, it is worth noting that both the Factor Graph Toolbox and Ceres Solver are tools for solving optimization problems, particularly those involving graphical models. Graphical models are a way to represent complex probabilistic relationships between variables, and they often arise in machine learning, computer vision, and other fields.

Therefore, it is possible that the Factor Graph Toolbox and Ceres Solver share some underlying algorithms or techniques for solving optimization problems on graphical models, which could explain why their output appears similar. However, without access to the implementation details of the optimize function, it is difficult to say for sure.

[Justin]

I realize this thread is old, but here's some updated info for those that stumble across this like I did when I had similar questions. As of July 2nd, 2025 (R2025a), the online documentation states that optimize does use Ceres Solver:

https://www.mathworks.com/help/nav/ref/factorgraph.optimize.html

The optimize function optimizes a factor graph to find a solution that minimizes the cost of the nonlinear least squares problem formulated by the factor graph. The factor graph optimization utilizes the Ceres Solver for node state covariance estimation, a process that incurs higher computation costs and longer estimation times as the number of nodes increases. For more information about Ceres-Solver covariance estimation, see http://ceres-solver.org/nnls_covariance.html.