This commit is contained in:
Jincheng Chen 2021-11-16 07:42:58 +00:00
Родитель c7d4cc042b
Коммит 4e97e6d893
3 изменённых файлов: 7 добавлений и 5 удалений

Просмотреть файл

@ -12,8 +12,10 @@ In the repo, user could learn about:
## Get Started and Documentation
- [Customized parser tutorial](./docs/Customized-Parser-Tutorial.rst).
- [Transform tutorial](./docs/Transform-Tutorial.rst).
- [Installation guide](./docs/Installation-Guide.rst)
- [Customized parser tutorial](./docs/Customized-Parser-Tutorial.rst)
- [Transform tutorial](./docs/Transform-Tutorial.rst)
- [FreeForm2 language spec](./docs/FreeForm2-Language.rst)
## Contributing

Просмотреть файл

@ -10,4 +10,4 @@ The basic methodology is to register custom parser to parser factory via reflect
- See `Installation Guide <./Installation-Guide.rst>`__ for dependencies and your parser installation.
- See `Transform tutorial <./Transform-tutorial.rst>`__ for how to use LightGBM integrated with your parser version.
- See `Transform tutorial <./Transform-Tutorial.rst>`__ for how to use LightGBM integrated with your parser version.

Просмотреть файл

@ -1,7 +1,7 @@
LightGBM Transform Tutorial
===========================
The purpose of this document is to give you a tutorial on how to do transformation in LightGBM with `FreeForm2Parser <./examples/freeform2_parser.cpp>`__.
The purpose of this document is to give you a tutorial on how to do transformation in LightGBM with `FreeForm2Parser <../examples/freeform2_parser.cpp>`__.
Transformation is a process to convert data/feature from one format to another.
Now we support two kinds of transformations in LightGBM:
@ -42,7 +42,7 @@ Data preparation
**Note**: transform value is the content of transform file.
Transform file is not a supplement of raw features, but all used for training. Use "Linear" type if you want to keep the original ones.
See `FreeForm2 language spec <./FreeForm2-Language>`__ and learn more about the grammar.
See `FreeForm2 language spec <./FreeForm2-Language.rst>`__ and learn more about the grammar.
**Note**: transformed feature index ranges from 0 to the maximum "Input" value given in transform file.
By default, will pad 0 as feature value for missing indices within the range.