def analyse(sm, labels, root='', plotting_context=None, file_format='pdf',
force_silhouette=False, threshold=None):
"""Perform analysis.
Parameters
----------
sm : array, shape = [n_samples, n_samples]
Precomputed similarity matrix.
labels : array, shape = [n_samples]
Association of each sample to a clsuter.
root : string
The root path for the output creation.
plotting_context : dict, None, or one of {paper, notebook, talk, poster}
See seaborn.set_context().
file_format : ('pdf', 'png')
Choose the extension for output images.
"""
sns.set_context(plotting_context)
if force_silhouette or sm.shape[0] < 8000:
silhouette.plot_clusters_silhouette(1. - sm.toarray(), labels,
max(labels), root=root,
file_format=file_format)
else:
logging.warn(
"Silhouette analysis is not performed due to the "
"matrix dimensions. With a matrix %ix%i, you would need to "
"allocate %.2fMB in memory. If you know what you are doing, "
"specify 'force_silhouette = True' in the config file in %s, "
"then re-execute the analysis.\n", sm.shape[0], sm.shape[0],
sm.shape[0]**2 * 8 / (2.**20), root)
# Generate dendrogram
import scipy.spatial.distance as ssd
Z = hierarchy.linkage(ssd.squareform(1. - sm.toarray()), method='complete',
metric='euclidean')
plt.close()
fig, (ax) = plt.subplots(1, 1)
fig.set_size_inches(20, 15)
hierarchy.dendrogram(Z, ax=ax)
ax.axhline(threshold, color="red", linestyle="--")
plt.show()
filename = os.path.join(root, 'dendrogram_{}.{}'
.format(extra.get_time(), file_format))
fig.savefig(filename)
logging.info('Figured saved %s', filename)
plt.close()
fig, (ax) = plt.subplots(1, 1)
fig.set_size_inches(20, 15)
plt.hist(1. - sm.toarray(), bins=50, normed=False)
plt.ylim([0, 10])
fig.savefig(filename + "_histogram_distances.pdf")
评论列表
文章目录
