PGBM

Probabilistic Gradient Boosting Machines (PGBM) is a probabilistic gradient boosting framework in Python based on PyTorch/Numba, developed by Airlab in Amsterdam. It provides the following advantages over existing frameworks:

• Probabilistic regression estimates instead of only point estimates. (example)
• Auto-differentiation of custom loss functions. (example, example)
• Native (multi-)GPU-acceleration. (example, example)
• Ability to optimize probabilistic estimates after training for a set of common distributions, without retraining the model. (example)

It is aimed at users interested in solving large-scale tabular probabilistic regression problems, such as probabilistic time series forecasting. For more details, read our paper or check out the examples.

Installation

Run pip install pgbm from a terminal within a Python (virtual) environment of your choice.

Verification

• Download & run an example from the examples folder to verify the installation is correct:
  • Run this example to verify ability to train & predict on CPU with Torch backend.
  • Run this example to verify ability to train & predict on GPU with Torch backend.
  • Run this example to verify ability to train & predict on CPU with Numba backend.
• Note that when training on the GPU, the custom CUDA kernel will be JIT-compiled when initializing a model. Hence, the first time you train a model on the GPU it can take a bit longer, as PGBM needs to compile the CUDA kernel.
• When using the Numba-backend, several functions need to be JIT-compiled. Hence, the first time you train a model using this backend it can take a bit longer.
• To run the examples some additional packages such as scikit-learn or matplotlib are required; these should be installed separately via pip or conda.

Dependencies

The core package has the following dependencies which should be installed separately (installing the core package via pip will not automatically install these dependencies).

Torch backend

• CUDA Toolkit matching your PyTorch distribution (https://developer.nvidia.com/cuda-toolkit)
• PyTorch >= 1.7.0, with CUDA 11.0 for GPU acceleration (https://pytorch.org/get-started/locally/). Verify that PyTorch can find a cuda device on your machine by checking whether torch.cuda.is_available() returns True after installing PyTorch.
• PGBM uses a custom CUDA kernel which needs to be compiled, which may require installing a suitable compiler. Installing PyTorch and the full CUDA Toolkit should be sufficient, but open an issue if you find it still not working even after installing these dependencies.

Numba backend

• Numba >= 0.53.1 (https://numba.readthedocs.io/en/stable/user/installing.html).

The Numba backend does not support differentiable loss functions and GPU training is also not supported using this backend.

Support

See the examples folder for examples, an overview of hyperparameters and a function reference. In general, PGBM works similar to existing gradient boosting packages such as LightGBM or xgboost (and it should be possible to more or less use it as a drop-in replacement), except that it is required to explicitly define a loss function and loss metric.

In case further support is required, open an issue.

Reference

Olivier Sprangers, Sebastian Schelter, Maarten de Rijke. Probabilistic Gradient Boosting Machines for Large-Scale Probabilistic Regression. Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 21), August 14–18, 2021, Virtual Event, Singapore.

The experiments from our paper can be replicated by running the scripts in the experiments folder. Datasets are downloaded when needed in the experiments except for higgs and m5, which should be pre-downloaded and saved to the datasets folder (Higgs) and to datasets/m5 (m5).

License

This project is licensed under the terms of the Apache 2.0 license.

Acknowledgements

This project was developed by Airlab Amsterdam.