sandy4321 / slir

Python package for Sparse Linear Regression (SLiR)

SLiR: Sparse Linear Regression

Sparse linear regression (SLiR), developed by Misato Tanaka at ATR.

Original Sparse Linear Regerssion toolbox for Matlab is available at http://www.cns.atr.jp/cbi/sparse_estimation/sato/VBSR.html.

Installation

Run the following command:

$ pip install git+https://github.com/KamitaniLab/slir.git

Usage

import slir

model = slir.SparseLinearRegression(n_iter=100)

model.fit(x, y)
y_pred = model.predict(x_test)

• x, x_test: numpy array of training and test input features
• y: target vector

The API of this function is compatible with the regression in scikit-learn. For demonstration, try demo_slir.py.

References

Sato M. (2001) On-line model selection based on the variational Bayes. Neural Computation, 13, 1649-1681. http://www.mitpressjournals.org/doi/abs/10.1162/089976601750265045

License

The scripts provided here are released under the MIT license (http://opensource.org/licenses/mit-license.php).