[R] Document handling of indexes #10019
Conversation
doc/contrib/coding_guide.rst
Outdated
| - Categorical features are defined in R as a ``factor`` type which encodes with base-1 indexing. When categorical features are passed as R ``factor`` types, the conversion is done automatically to base-0 indexing, but if the user whishes to manually supply categorical features as already-encoded integers, then those integers need to already be in base-0 encoding. | ||
| - Categorical labels for DMatrices do not undergo any extra processing - the user must supply base-0 encoded labels. | ||
| - A function to retrieve class-specific coefficients when using the linear coefficients history callback takes a class index parameter, which also does not undergo any conversion (i.e. user must pass a base-0 index), in order to match with the label logic - that is, the same class index will refer to the class encoded with that number in the DMatrix ``label`` field. | ||
|
|
There was a problem hiding this comment.
How about evaluation output:
[0]: rmse ...
[1]: rmse ...
doc/contrib/coding_guide.rst
Outdated
| - Categorical labels for DMatrices do not undergo any extra processing - the user must supply base-0 encoded labels. | ||
| - A function to retrieve class-specific coefficients when using the linear coefficients history callback takes a class index parameter, which also does not undergo any conversion (i.e. user must pass a base-0 index), in order to match with the label logic - that is, the same class index will refer to the class encoded with that number in the DMatrix ``label`` field. | ||
|
|
||
| New additions to the R interface that take on indexable elements should be mindful of these conventions and try to mimic R's behavior as much as possible. |
There was a problem hiding this comment.
How about the feature index from the tree plot and JSON model? I think for the JSON model, the output will continue to be 0-based, not sure about what to do with the plot. In addition, the categorical features in tree dump and plot.
There was a problem hiding this comment.
Added, alongside with other cases that were missing.
Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
Do you suggest to move this to some other file? |
|
Yes, perhaps somewhere as part of the R document like a vignette since it's quite R-specific; otherwise, creating a new file like |
Moved it to a new file under Also added a few notes about base-0 encodings in the DMatrix docs. |
|
@trivialfis @hcho3 Looks like the docs didn't render correctly, and didn't include this page in the index. I am not familiar with RST format. Could you look into it? |
|
I will look into it tomorrow night. |
|
@trivialfis I am looking at this now. |
|
@david-cortes #10030 should fix the doc. |
ref #9810
This PR adds some notes about handling of index numbers in R, as requested in previous comment.