Closed frousseu closed 11 months ago
Any thoughts about using maxnet instead of maxent.jar
? The differences are covered in the paper (https://onlinelibrary.wiley.com/doi/10.1111/ecog.03049), and the results are about the same, with no need for an extra java (closed-source) dependency?
From my limited experience, maxent.jar seems faster and less buggy than maxnet, but I agree maxnet is more open-source and has fewer dpendencies. I changed the algorithm to maxent.jar, because it scales variables behing the scene, while maxnet does not and I was running into numerical problems resulting from non-scaled variables. We could easily scale variables ourselves, but it can be a bit of a pain when exploring results and I wanted to avoid that. Another option could be to wait for maxnet to scale variables https://github.com/mrmaxent/maxnet/issues/24 (or to submit a pull request to maxnet to allow for scaling in glmnet).
Described a bit more some arguments and functions, simplified the create background function and added the thickening and inclusion_buffer method which were not working/available, now accepts the - in soilgrids variable names, reduced the resolution for faster runs