This repository provides the Python module rfflearn which is a Python library of random Fourier features (hereinafter abbreviated as RFF) [1, 2] for kernel methods, like support vector machine [3, 4] and Gaussian process model [5]. Features of this module are:

• User-friendly interfaces: Interfaces of the rfflearn module are quite close to the scikit-learn library,
• Example code first: This repository provides plenty of example code to demonstrate that RFF is useful for actual machine learning tasks,
• GPU support: Some classes in the rfflearn module provides both GPU training and inference for faster computation,
• Wrapper to the other library: Interface to optuna and SHAP are provided for easier hyperparameter tuning and feature importance analysis.

Now, this module supports the following methods:

RFF can be applicable to many other machine learning algorithms than the above. The author will provide implementations of the other algorithms soon.

NOTE: The author confirmed that the rfflearn module works on PyTorch 2.0! 🎉

Interfaces provided by the module rfflearn are quite close to scikit-learn. For example, the following Python code is a sample usage of RFFSVC (support vector machine with random Fourier features) class.

>>> import numpy as np
>>> import rfflearn.cpu as rfflearn                     # Import module
>>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]])  # Define input data
>>> y = np.array([1, 1, 2, 2])                          # Defile label data
>>> svc = rfflearn.RFFSVC().fit(X, y)                   # Training (on CPU)
>>> svc.score(X, y)                                     # Inference (on CPU)
1.0
>>> svc.predict(np.array([[-0.8, -1]]))
array([1])

This module supports training/inference on GPU. For example, the following Python code is a sample usage of RFFGPC (Gaussian process classifier with random Fourier features) on GPU. The following code requires PyTorch (>= 1.7.0).

>>> import numpy as np
>>> import rfflearn.gpu as rfflearn                     # Import module
>>> X = np.array([[-1, -1], [-2, -1], [1, 1], [2, 1]])  # Define input data
>>> y = np.array([1, 1, 2, 2])                          # Defile label data
>>> gpc = rfflearn.RFFGPC().fit(X, y)                   # Training on GPU
>>> gpc.score(X, y)                                     # Inference on GPU
1.0
>>> gpc.predict(np.array([[-0.8, -1]]))
array([1])

See the examples directory for more detailed examples.

The author tried SVC (support vector classifier) and GPC (Gaussian process classifier) with RFF to the MNIST dataset which is one of the famous benchmark datasets on the image classification task, and the author has got better performance and much faster inference speed than kernel SVM. The following table gives a brief comparison of kernel SVM, SVM with RFF, and GPC with RFF. See the example of RFF SVC module and RFF GP module for more details.

This module also has interfaces to some feature importance methods, like SHAP [6] and permutation importance [7]. The author tried SHAP and permutation importance to RFFGPR trained on the California housing dataset, and the followings are the visualization results obtained by rfflearn.shap_feature_importance and rfflearn.permutation_feature_importance.

The author recommends using a docker image for building the environment for the rfflearn, however, of course, you can install the necessary packages on your environment directly. See SETUP.md for more details.

• The name of this module is changed from pyrff to rfflearn on Oct 2020, because the package name pyrff already exists in PyPI.
• If the number of training data is huge, an error message like RuntimeError: The task could not be sent to the workers as it is too large for 'send_bytes' will be raised from the joblib library. The reason for this error is that the sklearn.svm.LinearSVC uses joblib as a multiprocessing backend, but joblib cannot deal huge size of the array which cannot be managed with 32-bit address space. In this case, please try n_jobs = 1 option for the RFFSVC or ORFSVC function. Default settings are n_jobs = -1 which means automatically detecting available CPUs and using them. (This bug information was reported by Mr. Katsuya Terahata @ Toyota Research Institute Advanced Development. Thank you so much for the reporting!)
• Application of RFF to the Gaussian process model is not straightforward. See this document for mathematical details.

MIT Licence

[1] A. Rahimi and B. Recht, "Random Features for Large-Scale Kernel Machines", NIPS, 2007. PDF

[2] F. X. Yu, A. T. Suresh, K. Choromanski, D. Holtmann-Rice and S. Kumar, "Orthogonal Random Features", NIPS, 2016. PDF

[3] V. Vapnik and A. Lerner, "Pattern recognition using generalized portrait method", Automation and Remote Control, vol. 24, 1963.

[4] B. Boser, I. Guyon and V. Vapnik, "A training algorithm for optimal margin classifiers", COLT, pp. 144-152, 1992 URL

[5] C. Rasmussen and C. Williams, "Gaussian Processes for Machine Learning", MIT Press, 2006.

[6] S. M. Lundberg and S. Lee, "A Unified Approach to Interpreting Model Predictions", NIPS, 2017. PDF

[7] L. Breiman, "Random Forests", Machine Learning, vol. 45, pp. 5-32, Springer, 2001. Springer website.

Tetsuya Ishikawa (EMail, Website)