Fern-based classifiers

In a nutshell, a fern is a decision tree where all nodes of each depth level use the same feature and split criteria. Although it is simpler than a tree, when used as an ensemble of ferns, it turns fast and powerful!

Random Ferns were first proposed in [1] for image keypoint recognition tasks. Later, in [2] a general purpose machine learning method was developed in R, and is the base line of this study.

Our implementations follow the python scikit-learn API.

Authors

This implementation was developed and maintained by Mario Juez-Gil from ADMIRABLE research group of the University of Burgos, with the help and useful advice from Álvar Arnaiz-González, Juan J. Rodriguez, and Ludmila Kuncheva. The work was supervised by César García-Osorio and Carlos López-Nozal.

Usage

Cython was used for speeding the algorithms. Because of that, the code must be compiled as follows:

python setup.py build_ext --inplace

Once compiled, it can be used like any scikit-learn classifier. A basic example could be:

example.py

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix
from fc.ferns import FernClassifier

iris = load_iris()
X_train, X_test, y_train, y_test = train_test_split(iris.data, iris.target,
                                                    test_size=0.5)
fern = FernClassifier(depth=5, random_state=1)
fern.fit(X_train, y_train)
y_pred = fern.predict(X_test)
print(confusion_matrix(y_test, y_pred))

output:

[[24  1  0]
 [ 3 19  1]
 [ 0  7 20]]

References

[1] M. Ozuysal, M. Calonder, V. Lepetit, and P. Fua, “Fast Keypoint Recognition Using Random Ferns,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 32, no. 3, pp. 448–461, Mar. 2010. DOI: 10.1109/TPAMI.2009.23

[2] M. B. Kursa, “rFerns: An Implementation of the Random Ferns Method for General-Purpose Machine Learning,” J. Stat. Softw., vol. 61, no. 10, pp. 1–13, Nov. 2014. DOI: 10.18637/jss.v061.i10

License

Licensed under the GNU GPLv3, please see the LICENSE file for more details.

Acknowledgements

This work was partially supported by the Consejería de Educación of the Junta de Castilla y León and by the European Social Fund with the EDU/1100/2017 pre-doctoral grants; by the project TIN2015-67534-P (MINECO/FEDER, UE) of the Ministerio de Economía Competitividad of the Spanish Government and the project BU085P17 (JCyL/FEDER, UE) of the Junta de Castilla y León both cofinanced from European Union FEDER funds.