When using parametric distance models it is often the case that the fitting fails due to a singular matrix error or because the SVD did not converge.
One possible solution to this is to use the multistart fit option and sample a number of starting positions, at a cost of increased fit time. However, currently if any single starting paramter fails the whole fit fails.
Proposal
A try, except block is added in the multi-start function with failed fits skiped. A warning should then be raised if any starting conditions are skipped.
Furthermore, this could be parallelised to speed up the perfomance and give a progress bar.
When using parametric distance models it is often the case that the fitting fails due to a singular matrix error or because the SVD did not converge.
One possible solution to this is to use the
multistartfit option and sample a number of starting positions, at a cost of increased fit time. However, currently if any single starting paramter fails the whole fit fails.Proposal
A try, except block is added in the multi-start function with failed fits skiped. A warning should then be raised if any starting conditions are skipped.
Furthermore, this could be parallelised to speed up the perfomance and give a progress bar.