How to stepwise with Biogeme

[Karine Xu]

Dear Professor Bierlaire,

I wonder if can I do the forward/backward stepwise of the hybrid choice model with Biogeme. Could you please let me know the relative algorithm and do you have some examples? Thank you very much!

Kind regards,

Chenchen

[xinyu zuo]

Hello everyone, 

       Hope everything is well with you. 

       Have you solved it? I also have a similar question. If I use the same model, the same dataset, and the same standardization method, but only modify the utility function in the code. Can I calibrate the parameters step by step or in stages using Biogeme? Will this lead to inconsistent parameters? (Because I have too many parameters, the overall calculation takes a very long time, even more than two weeks.) 

       I just tried it, and there are indeed some differences in the parameter values in the results.What should I do？

[Michel Bierlaire]

Yes, you can use sequential estimation to speed up the process. This is described here:
https://transp-or.epfl.ch/documents/technicalReports/Bier18b.pdf
The sequential estimator is consistent but not efficient.

Also, during the model development, it is OK to decrease the number of draws, to speed up the estimation time. Once you are happy with the specification, use a computing server, with a lot of parallel processors, to estimate the final model with a high number of draws.

> --
> You received this message because you are subscribed to the Google Groups "Biogeme" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to biogeme+u...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/biogeme/c4e55d5f-2166-49fd-bb5e-9e69950af23bn%40googlegroups.com.

Michel Bierlaire
Transport and Mobility Laboratory
School of Architecture, Civil and Environmental Engineering
EPFL - Ecole Polytechnique Fédérale de Lausanne
http://transp-or.epfl.ch
http://people.epfl.ch/michel.bierlaire