if verbose:
print galaxy_ids[out]
np.savetxt(base_dir + 'data/outliers.txt', galaxy_ids[out])
if load_pca:
print 'Loading PCA...'
rpca = cPickle.load(open(base_dir + 'data/DCT_PCA.pickle', 'rb'))
X = rpca.transform(X)
else:
# do the PCA
if verbose:
print 'Doing PCA...'
rpca = RandomizedPCA(n_components=npca, copy=False)
X = rpca.fit_transform(X)
rpca.galaxy_ids = galaxy_ids[notout]
cPickle.dump(rpca, open(base_dir + 'data/DCT_PCA.pickle', 'wb'))
np.save(base_dir + 'data/PCA_dist_no_outliers', X)
plt.plot(rpca.explained_variance_ratio_.cumsum())
plt.ylabel('Cumulative Fractional Explained Variance')
plt.xlabel('Number of Components')
plt.savefig(plot_dir + 'explained_variance.png')
if doshow:
plt.show()
# now plot after removing outliers
fig = plot_pc_projections(X, npca=6)
fig.savefig(plot_dir + 'PC_dist_no_outliers.png')
if doshow:
plt.show()
