TinyGBT

TinyGBT(Tiny Gradient Boosted Trees) is a 200 line gradient boosted trees implementation written in pure python.

Since this code is not for production, it is not optimized for speed and memory usage.

Experiment

• Data: LightGBM's regression example data
• TinyGBT is quite slower than LightGBM, but achieve almost same testset RMSE on similar parameter settings.

| - | LightGBM | TinyGBT | | --- | --- | --- | | RMSE of TestSet | 0.45652 | 0.45934 |

Reproduce experiment

• TinyGBT

git clone https://github.com/lancifollia/tinygbt.git
cd tinygbt
python example.py

• LightGBM
  • run this code

Features

• For now, Regression with L2 loss supported only.

References

• [1] T. Chen and C. Guestrin. XGBoost: A Scalable Tree Boosting System. 2016.
• [2] G. Ke et al. LightGBM: A Highly Efficient Gradient Boosting Decision Tree. 2017.

License

MIT