fitcsvm not converging with large dataset

Hello,

This will be a very vague question so I don't suppose it can really be answered. After hours of trying to solve the issue I must give it a try..

I have 3 fitcsvm (each for one column) running on a dataset (array) 3012000x4. When I do not add noise to the data I have no issues - all SVM converges.

If noise is added the convergece depends on the size of the dataset. If I run fitcsvm using the whole dataset - two SVM converge & one does not converge. If use a reduced daset (e.g. 2500000x4) - all SVMs converge.

Any ideas on what I can look into to solve this issue?

Thank you,

[Navya Seelam]

This might be because of the overfitting in SVM. Did you try changing the BoxConstraint parameter of fitcsvm to prevent overfit.

[Aleksander Tyczynski]

Thank you, I will have a look at this and come back to you if it solves the issue. :)

[Aleksander Tyczynski]

I changed the 'IterationLimit' by an order of magnitude and it now converges. Thank you