fabianp

"""
This module implements the Lowess function for nonparametric regression.

Functions:
lowess Fit a smooth nonparametric regression curve to a scatterplot.

For more information, see

William S. Cleveland: "Robust locally weighted regression and smoothing
scatterplots", Journal of the American Statistical Association, December 1979,

fabianp

=====================
Lasso and Elastic Net
=====================

Lasso and elastic net (L1 and L2 penalisation) implemented using a
coordinate descent.

The coefficients can be forced to be positive.

fabianp

fabianp

import numpy as np
from scipy import linalg
from datetime import datetime

import gc

mu_sec = 1e-6 # number of seconds in one microseconds

solve_time =[]
solve_triangular_time =[]

fabianp

"""
Just some experiments with pls
"""

import numpy as np

m, n = 3, 3
X = np.random.randn(m, n)
Y = np.random.randn(m, n)

fabianp

import pylab as pl
import numpy as np
import scipy.linalg

nrm2, = scipy.linalg.get_blas_funcs(('nrm2',), np.empty(0, dtype=np.float32))

x_min, x_max = -4., 0
y_min, y_max = 0, 4.
h = 134567e-7

fabianp

import numpy as np

from scikits.learn import svm, grid_search, datasets
iris = datasets.load_iris()
parameters = {'C': [np.array(1)]}
svr = svm.SVC()
clf = grid_search.GridSearchCV(svr, parameters)
clf.fit(iris.data, iris.target)
print clf.grid_points_scores_

fabianp

"""
===============================================
Handwritten digits and Locally Linear Embedding 
===============================================

An illustration of locally linear embedding on the digits dataset.
"""

# Author: Fabian Pedregosa -- <fabian.pedregosa@inria.fr>
# License: BSD, (C) INRIA 2011

fabianp

import numpy as np
from scipy import linalg

def cgs(A):
    """Classical Gram-Schmidt (CGS) algorithm"""
    m, n = A.shape
    R = np.zeros((n, n))
    Q = np.empty((m, n))
    R[0, 0] = linalg.norm(A[:, 0])
    Q[:, 0] = A[:, 0] / R[0, 0]

fabianp

import numpy as np
from scipy import linalg

def ridge(A, b, alphas):
    """Return coefficients for regularized least squares 
               ||A x - b|| + alpha ||x||
    """ 
    U, s, V = linalg.svd(X, full_matrices=False)
    d = np.dot(U.T, y) / (s + alphas[:, np.newaxis] / s)
    return np.dot(d, V)