5 unstable releases

0.3.0 Mar 14, 2022
0.2.2 Feb 22, 2022
0.2.1 Jan 13, 2022
0.2.0 Jan 11, 2022
0.1.0 Dec 30, 2021

#395 in Machine learning

Download history 12/week @ 2023-10-28 6/week @ 2023-11-04 5/week @ 2023-11-11 12/week @ 2023-11-18 44/week @ 2023-11-25 54/week @ 2023-12-02 67/week @ 2023-12-09 4/week @ 2023-12-16 15/week @ 2023-12-23 70/week @ 2023-12-30 491/week @ 2024-01-06 8/week @ 2024-01-13 9/week @ 2024-01-20 23/week @ 2024-01-27 4/week @ 2024-02-03 24/week @ 2024-02-10

61 downloads per month
Used in changeforest

BSD-3-Clause

45KB
991 lines

Biosphere

Simple, fast random forests.

Random forests with a runtime of O(n d log(n) + n_estimators d n max_depth) instead of O(n_estimators mtry n log(n) max_depth).

biosphere is available as a rust crate and as a Python package.

Benchmarks

Ran on an M1 Pro with n_jobs=4. Wall-time to fit a Random Forest including OOB score with 400 trees to the NYC Taxi dataset, minimum over 10 runs. After feature engineering, the dataset consists of 5 numerical and 7 one-hot encoded features.

model 1000 2000 4000 8000 16000 32000 64000 128000 256000 512000 1024000 2048000
biosphere 0.04s 0.08s 0.15s 0.32s 0.65s 1.40s 2.97s 6.48s 15.53s 37.91s 96.69s 231.82s
scikit-learn 0.28s 0.34s 0.46s 0.69s 1.23s 2.47s 4.99s 10.49s 22.11s 51.04s 118.95s 271.03s

Dependencies

~3MB
~55K SLoC