search

dmlc / xgboost

Scalable, Portable and Distributed Gradient Boosting (GBDT, GBRT or GBM) Library, for Python, R, Java, Scala, C++ and more. Runs on single machine, Hadoop, Spark, Dask, Flink and DataFlow
https://xgboost.readthedocs.io/en/stable/
Apache License 2.0
26.16k stars 8.71k forks source link

Setting bias to zero in gblinear #8848

Open jaguerrerod opened 1 year ago

jaguerrerod commented 1 year ago

How can I set the bias to 0 in gblinear booster now? And what happened with lambda_bias? (a way to force bias to zero) Documentation about gblinear is scarce or I don't know where looking for.

trivialfis commented 1 year ago

It's not supported.

jaguerrerod commented 1 year ago

So there isn't a way to force a linear model without offset?

trivialfis commented 1 year ago

No, we spend most of the time on tree models and the linear model is less attended.

jaguerrerod commented 1 year ago

I set base_score to 0 and getting very small values of offset. My dataset is centered at 0 (mean response is zero). Could this works in general? How is computed the offset in gblinear?