baohan
/
scikit-learn
mirror of https://github.com/scikit-learn/scikit-learn.git
0
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
scikit-learn/examples/linear_model/plot_lasso_lars.py

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

39 lines
1011 B
Python
Raw Normal View History

Add a pure-python version of LARS and refactor structure in glm. Divided the glm module in submodules, although API should not be affected since methods are imported from the __init__method. The lars method is a work in progress. For now, it is several times slower than the C version, but it is less buggy and implements the lasso variant.
2010-08-31 20:43:45 +08:00
"""
New implementation of LARS algorithm. Small performance improvements (~10%) and it no longer uses custom dot matrix function dot_over. Except for solve_triangular, all other methods can operate on float32 arrays. Thinkint out loud: The complexity of LARS is the same as an OLS, and looking at the benchmarks scikits/learn/glm/benchmarks/bench_glm.py it shows that we are faster than scipy's lstsq, so I don't think there is much more room for improvement. I'll benchmark against the R package and see how far we are. In arrayfuncs: removed dot_over and added min_pos.
2010-11-05 21:54:09 +08:00
=====================
Lasso path using LARS
=====================
Add a pure-python version of LARS and refactor structure in glm. Divided the glm module in submodules, although API should not be affected since methods are imported from the __init__method. The lars method is a work in progress. For now, it is several times slower than the C version, but it is less buggy and implements the lasso variant.
2010-08-31 20:43:45 +08:00
New implementation of LARS algorithm. Small performance improvements (~10%) and it no longer uses custom dot matrix function dot_over. Except for solve_triangular, all other methods can operate on float32 arrays. Thinkint out loud: The complexity of LARS is the same as an OLS, and looking at the benchmarks scikits/learn/glm/benchmarks/bench_glm.py it shows that we are faster than scipy's lstsq, so I don't think there is much more room for improvement. I'll benchmark against the R package and see how far we are. In arrayfuncs: removed dot_over and added min_pos.
2010-11-05 21:54:09 +08:00
Computes Lasso Path along the regularization parameter using the LARS
DOC spellfixes
2013-06-27 21:09:16 +08:00
algorithm on the diabetes dataset. Each color represents a different
Some doc for the example "Lasso path using LARS"
2012-05-24 21:41:40 +08:00
feature of the coefficient vector, and this is displayed as a function
of the regularization parameter.
pretifying the LAR / LARS examples to match with results on wikipedia page
2010-09-21 17:00:24 +08:00
Add a pure-python version of LARS and refactor structure in glm. Divided the glm module in submodules, although API should not be affected since methods are imported from the __init__method. The lars method is a work in progress. For now, it is several times slower than the C version, but it is less buggy and implements the lasso variant.
2010-08-31 20:43:45 +08:00
"""
# Author: Fabian Pedregosa <fabian.pedregosa@inria.fr>
ENH : more work on LARS (doc + examples)
2010-09-01 00:37:41 +08:00
# Alexandre Gramfort <alexandre.gramfort@inria.fr>
DOC: Replaced all BSD style licenses with "BSD 3 clause" Replaced all BSD style licenses with "BSD 3 clause" Not checked yet! Removed duplicate "3 clause, 3 clause" Removed trailing period if exists Fixed some missed licences, still about 50 to do, but those can be automated Think I got the last of them. Apparently me and sed have different ideas of regex. Found a few more
2013-04-30 14:23:46 +08:00
# License: BSD 3 clause
Add a pure-python version of LARS and refactor structure in glm. Divided the glm module in submodules, although API should not be affected since methods are imported from the __init__method. The lars method is a work in progress. For now, it is several times slower than the C version, but it is less buggy and implements the lasso variant.
2010-08-31 20:43:45 +08:00
import numpy as np
Updated examples to use pyplot for plotting instead of pylab Some examples not updated because scripts throw error (probably not because of the change)
2014-05-15 04:31:03 +08:00
import matplotlib.pyplot as plt
Add a pure-python version of LARS and refactor structure in glm. Divided the glm module in submodules, although API should not be affected since methods are imported from the __init__method. The lars method is a work in progress. For now, it is several times slower than the C version, but it is less buggy and implements the lasso variant.
2010-08-31 20:43:45 +08:00
Move project directory from scikits.learn to sklearn
2011-09-02 17:00:02 +08:00
from sklearn import linear_model
from sklearn import datasets
Add a pure-python version of LARS and refactor structure in glm. Divided the glm module in submodules, although API should not be affected since methods are imported from the __init__method. The lars method is a work in progress. For now, it is several times slower than the C version, but it is less buggy and implements the lasso variant.
2010-08-31 20:43:45 +08:00
DOC return_X_y for load diabetes (#14762)
2019-08-25 11:17:01 +08:00
X, y = datasets.load_diabetes(return_X_y=True)
Add parameter precompute='auto' for *LARS classes. Also some refactoring.
2010-11-22 22:10:21 +08:00
P3K: use print as a function in the examples Also corrected some typos, including two people misspelling their own names :)
2013-02-01 22:04:03 +08:00
print("Computing regularization path using the LARS ...")
Remove unused variable alphas from the LARS example. (#9882)
2017-10-07 23:04:08 +08:00
_, _, coefs = linear_model.lars_path(X, y, method="lasso", verbose=True)
ENH: make the LassoLARS and LassoCD path examples easier to compare
2010-12-09 16:32:32 +08:00
revert the plot style to the LARS paper conventions
2010-12-10 01:57:45 +08:00
xx = np.sum(np.abs(coefs.T), axis=1)
xx /= xx[-1]
Add a pure-python version of LARS and refactor structure in glm. Divided the glm module in submodules, although API should not be affected since methods are imported from the __init__method. The lars method is a work in progress. For now, it is several times slower than the C version, but it is less buggy and implements the lasso variant.
2010-08-31 20:43:45 +08:00
Updated examples to use pyplot for plotting instead of pylab Some examples not updated because scripts throw error (probably not because of the change)
2014-05-15 04:31:03 +08:00
plt.plot(xx, coefs.T)
ymin, ymax = plt.ylim()
plt.vlines(xx, ymin, ymax, linestyle="dashed")
plt.xlabel("|coef| / max|coef|")
plt.ylabel("Coefficients")
plt.title("LASSO Path")
plt.axis("tight")
plt.show()