Hello Daniel,
so, I am using the package for a supply chain application and it works great, thanks!
I then needed to use the package without data, with a fix, given covariance matrix and expected value vector.
The first thing I did was adapt the code for this case:
https://github.com/bernardokp/NORTA/blob/main/Norta3
Then, your procedure for finding a positive-definite matrix after the NORTA matrix was found was not working well in my case. For any perturbation I chose I got the error message "Cholesky factorization failed, starting over with a modified cov matrix.". What I did was use this code:
https://github.com/bernardokp/NORTA/blob/main/nearestPD
Now, in my NORTA implementation, I have this projection instead of the search method with lambda. It is fast, and it always find a solution.
Best,
Bernardo
Hello Daniel,
so, I am using the package for a supply chain application and it works great, thanks!
I then needed to use the package without data, with a fix, given covariance matrix and expected value vector.
The first thing I did was adapt the code for this case:
https://github.com/bernardokp/NORTA/blob/main/Norta3
Then, your procedure for finding a positive-definite matrix after the NORTA matrix was found was not working well in my case. For any perturbation I chose I got the error message "Cholesky factorization failed, starting over with a modified cov matrix.". What I did was use this code:
https://github.com/bernardokp/NORTA/blob/main/nearestPD
Now, in my NORTA implementation, I have this projection instead of the search method with lambda. It is fast, and it always find a solution.
Best,
Bernardo