Deep Dive into scikit-learn's HistGradientBoosting Classifier and Regressor

• Link to slides
• Link to youtube

Running benchmarks

0. Install anaconda

1. Setup environment

conda env create -f environment.yml
conda activate 2019-pydata-nyc-hist

2. Run benchmarks for each library

First run will download the HIGGS dataset which is 2.6 GB!

# This is the number of cores (no hyperthreading)
export OMP_NUM_THREADS=12
python bench.py sklearn
python bench.py catboost
python bench.py lightgbm
python bench.py xgboost