An implementation of the SuperLearner algorithm in Python based on scikit-learn.
Now updated for Python 3.5
pip install git+git://github.com/alexpkeil1/SuPyLearner.git
Run the following in terminal
git clone https://github.com/alexpkeil1/supylearner.git
cd supylearner
python setup.py install
untested
Example (from examples/typical_usage.py)
# typical usage
import supylearner as sl
from sklearn import datasets, svm, linear_model, neighbors, svm
import numpy as np
# generate dataset
np.random.seed(100)
X, y = datasets.make_friedman1(1000)
ols = linear_model.LinearRegression()
elnet = linear_model.ElasticNetCV(l1_ratio = .1)
ridge = linear_model.RidgeCV()
lars = linear_model.LarsCV()
lasso = linear_model.LassoCV()
nn = neighbors.KNeighborsRegressor()
svm1 = svm.SVR(kernel = 'rbf')
svm2 = svm.SVR(kernel = 'poly')
lib = [ols, elnet, ridge,lars, lasso, nn, svm1, svm2]
libnames = ["OLS", "ElasticNet", "Ridge", "LARS", "LASSO", "kNN", "SVM rbf", "SVM poly"]
sl_inst = sl.SuperLearner(lib, libnames, loss = "L2")
sl_inst.fit(X, y)
sl_inst.summarize() Cross-validated risk estimates for each estimator in the library:
[['OLS' '5.889258599506167']
['ElasticNet' '6.014918631168619']
['Ridge' '5.889234044241611']
['LARS' '5.869122063410273']
['LASSO' '5.866767295197982']
['kNN' '7.037900242493755']
['SVM rbf' '6.242369357877353']
['SVM poly' '15.520506952085686']]
Coefficients:
[['OLS' '0.6367696182311826']
['ElasticNet' '0.0']
['Ridge' '0.0']
['LARS' '0.0']
['LASSO' '0.0']
['kNN' '0.36323038176881733']
['SVM rbf' '0.0']
['SVM poly' '0.0']]
(Not cross-valided) estimated risk for SL: 5.33523373261
sl.cv_superlearner(sl_inst, X, y, K = 5) Cross-validated risk estimates for each estimator in the library and SuperLearner:
[['OLS' '5.889258599506167']
['ElasticNet' '6.014918631168618']
['Ridge' '5.889234044241612']
['LARS' '5.869122063410273']
['LASSO' '5.866767295197983']
['kNN' '7.037900242493755']
['SVM rbf' '6.242369357877353']
['SVM poly' '15.520506952085688']
['SuperLearner' '5.34087144876605']]