lsqnonlin for double summation cost function

Hello,

I would like to use lsqnonlin to solve the Bundle Adjustment Problem. As you can see the objective function for this problem is defined using a double summation. It it possible to use lsqnonlin to express this type of problem? It seems as though it can only optimize over a single set of residuals instead of a sum (or double sum) of them.

Thanks!

Answer 1

[Matt J]

No, your objective function is allowed to return a residual array of any dimension.

https://www.mathworks.com/help/optim/ug/lsqnonlin.html#buuhch7-fun

[ar]

Thank you - do you mind expanding on how that would help my case?

I'm currently approaching this problem by just returning m*n components in my residual function.

[Matt J]

That's fine. There's nothing more you need to do. The shape of your residual array output is absolutely not relevant to anything.