Awesome

PyPI - Downloads

</div>

Perpetual

PerpetualBooster is a gradient boosting machine (GBM) algorithm which doesn't need hyperparameter optimization unlike other GBM algorithms. Similar to AutoML libraries, it has a budget parameter. Increasing the budget parameter increases the predictive power of the algorithm and gives better results on unseen data. Start with a small budget (e.g. 1.0) and increase it (e.g. 2.0) once you are confident with your features. If you don't see any improvement with further increasing the budget, it means that you are already extracting the most predictive power out of your data.

Benchmark

Hyperparameter optimization usually takes 100 iterations with plain GBM algorithms. PerpetualBooster achieves the same accuracy in a single run. Thus, it achieves up to 100x speed-up at the same accuracy with different budget levels and with different datasets.

The following table summarizes the results for the California Housing dataset (regression):

Perpetual budget	LightGBM n_estimators	Perpetual mse	LightGBM mse	Speed-up wall time	Speed-up cpu time
1.0	100	0.192	0.192	54x	56x
1.5	300	0.188	0.188	59x	58x
2.1	1000	0.185	0.186	42x	41x

The following table summarizes the results for the Cover Types dataset (classification):

Perpetual budget	LightGBM n_estimators	Perpetual log loss	LightGBM log loss	Speed-up wall time	Speed-up cpu time
0.9	100	0.091	0.084	72x	78x

The results can be reproduced using the scripts in the examples folder.

PerpetualBooster is a GBM but behaves like AutoML so it is benchmarked also against AutoGluon (v1.2, best quality preset), the current leader in AutoML benchmark. Top 10 datasets with the most number of rows are selected from OpenML datasets. The results are summarized in the following table for regression tasks:

OpenML Task	Perpetual Training Duration	Perpetual Inference Duration	Perpetual RMSE	AutoGluon Training Duration	AutoGluon Inference Duration	AutoGluon RMSE
Airlines_DepDelay_10M	518	11.3	29.0	520	30.9 <td style="background-color:green;color:white;"> 28.8 </td>
bates_regr_100	3421	15.1 <td style="background-color:green;color:white;"> 1.084 </td>	OOM	OOM	OOM
BNG(libras_move)	1956	4.2 <td style="background-color:green;color:white;"> 2.51 </td>	1922	97.6	2.53
BNG(satellite_image)	334	1.6	0.731	337	10.0 <td style="background-color:green;color:white;"> 0.721 </td>
COMET_MC	44	1.0 <td style="background-color:green;color:white;"> 0.0615 </td>	47	5.0	0.0662
friedman1	275	4.2 <td style="background-color:green;color:white;"> 1.047 </td>	278	5.1	1.487
poker	38	0.6 <td style="background-color:green;color:white;"> 0.256 </td>	41	1.2	0.722
subset_higgs	868	10.6 <td style="background-color:green;color:white;"> 0.420 </td>	870	24.5	0.421
BNG(autoHorse)	107	1.1 <td style="background-color:green;color:white;"> 19.0 </td>	107	3.2	20.5
BNG(pbc)	48	0.6 <td style="background-color:green;color:white;"> 836.5 </td>	51	0.2	957.1
average	465	3.9	-	464	19.7	-

PerpetualBooster outperformed AutoGluon on 8 out of 10 datasets, training equally fast and inferring 5x faster. The results can be reproduced using the automlbenchmark fork here.

Usage

You can use the algorithm like in the example below. Check examples folders for both Rust and Python.

from perpetual import PerpetualBooster

model = PerpetualBooster(objective="SquaredLoss")
model.fit(X, y, budget=1.0)

Documentation

Documentation for the Python API can be found here and for the Rust API here.

Installation

The package can be installed directly from pypi:

pip install perpetual

Using conda-forge:

conda install conda-forge::perpetual

To use in a Rust project and to get the package from crates.io:

cargo add perpetual

Contribution

Contributions are welcome. Check CONTRIBUTING.md for the guideline.

Paper

PerpetualBooster prevents overfitting with a generalization algorithm. The paper is work-in-progress to explain how the algorithm works. Check our blog post for a high level introduction to the algorithm.