uber
CausalTreeRegressor with the paramters 'criterion=causal_mse' can't been traind when dataset is much larger than example's
Describe the bug CausalTreeRegressor with the paramters 'criterion=causal_mse' can't been trained when dataset is much larger than example's synthetic_data (e.g. X.shape[0] > 1000000). But CausalTreeRegressor with the paramters 'criterion=standard_mse' works.
Environment (please complete the following information):
- OS: Ubuntu
- Python Version: 3.7
-
pandas==1.3.5,scikit-learn==1.0.2,cython==0.29.30,numpy==1.21.6
Additional context causalml package was compiled from lastest source code
When you say it can't be trained, what exactly happens when you try to do it?
When the rows of datasets are much more than 1 million, It would spent much time to train.
@lmaors, could you try the latest version from the repository? @alexander-pv further improved CausalTreeRegressor in #557.
Note that CausalTreeRegressor uses scikit-learn's tree implementation in Cython, and it's expected to be slower than XGBoost or LightGBM that are implemented in C++.
