def test_outlier_weights():
rand = np.random.RandomState(0)
X = rand.multivariate_normal([0, 0], [[12, 6], [6, 5]], size=1000)
pca = PCA(2).fit(X)
def check_results(Estimator, n_outliers, noise_level, rtol):
i = rand.randint(0, 100, size=n_outliers)
j = rand.randint(0, 2, size=n_outliers)
X2 = X.copy()
X2[i, j] += noise_level * rand.randn(n_outliers)
W2 = np.ones_like(X2)
W2[i, j] = 1. / noise_level
pca2 = Estimator(2, **KWDS[Estimator]).fit(X2, weights=W2)
assert_columns_allclose_upto_sign(pca.components_.T,
pca2.components_.T,
rtol=rtol)
for (n_outliers, noise_level, rtol) in [(1, 20, 1E-3), (10, 10, 3E-2)]:
for Estimator in ESTIMATORS:
yield check_results, Estimator, n_outliers, noise_level, rtol
w0.shape, w.shape, df[weights.values.ravel() != 0].shape, df.shape
# In[9]:
from wpca import WPCA as PCA
# In[10]:
from sklearn.pipeline import Pipeline
# from sklearn.decomposition import PCA
from sklearn.preprocessing import Imputer, StandardScaler
imp = Imputer(strategy='mean')
scl = StandardScaler()
pca = PCA()
pipeline = Pipeline([
('imp', imp),
('scl', scl),
('pca', pca),
])
scaler_pipeline = Pipeline([
('imp', imp),
('scl', scl),
])
data_pca = pipeline.fit_transform(df)
_scaled = scaler_pipeline.transform(df)
# ### Explained variance
#
# How much of the variance in the data is explained by each successive component?
print('NumPy covariance matrix: \n%s' %np.cov(X_std.T))
#Perform eigendecomposition on covariance matrix
cov_mat = np.cov(X_std.T)
eig_vals, eig_vecs = np.linalg.eig(cov_mat)
print('Eigenvectors \n%s' %eig_vecs)
print('\nEigenvalues \n%s' %eig_vals)
# Visually confirm that the list is correctly sorted by decreasing eigenvalues
eig_pairs = [(np.abs(eig_vals[i]), eig_vecs[:,i]) for i in range(len(eig_vals))]
print('Eigenvalues in descending order:')
for i in eig_pairs:
print(i[0])
#Variable loadings
pca = PCA(n_components=2)
pca.fit_transform(df1)
i = np.identity(df1.shape[1])
coef = pca.transform(i)
loadings = pd.DataFrame(coef, columns=['PC-1', 'PC-2'], index=M.columns)
print (loadings.head(10))
#print the eigenvalues for the first two principle components
print (pca.explained_variance_ratio_)
#Explained variance
pca = PCA().fit(X_std)
plt.plot(np.cumsum(pca.explained_variance_ratio_))
plt.xlabel('number of components')
plt.ylabel('cumulative explained variance')
plt.show()
print('NumPy covariance matrix: \n%s' %np.cov(X_std.T))
#Perform eigendecomposition on covariance matrix
cov_mat = np.cov(X_std.T)
eig_vals, eig_vecs = np.linalg.eig(cov_mat)
print('Eigenvectors \n%s' %eig_vecs)
print('\nEigenvalues \n%s' %eig_vals)
# Visually confirm that the list is correctly sorted by decreasing eigenvalues
eig_pairs = [(np.abs(eig_vals[i]), eig_vecs[:,i]) for i in range(len(eig_vals))]
print('Eigenvalues in descending order:')
for i in eig_pairs:
print(i[0])
#Variable loadings
pca = PCA(n_components=2)
pca.fit_transform(df1)
i = np.identity(df1.shape[1])
coef = pca.transform(i)
loadings = pd.DataFrame(coef, columns=['PC-1', 'PC-2'], index=M.columns)
print loadings.head(10)
#print the eigenvalues for the first two principle components
print pca.explained_variance_ratio_
#Explained variance
pca = PCA().fit(X_std)
plt.plot(np.cumsum(pca.explained_variance_ratio_))
plt.xlabel('number of components')
plt.ylabel('cumulative explained variance')
plt.show()