| *.swp | |
| *.pyc | |
| *.png | |
| data/* | |
| build |
| # Sample single piped shell unix commands to estimate the top 10 popular DBpedia resources | |
| # by counting incoming links between matching wikipedia articles | |
| % time curl http://downloads.dbpedia.org/3.5/en/page_links_en.nt.bz2 \ | |
| | bzcat \ | |
| | head -1000000 \ | |
| | sed -e 's/.*\/\(.*\)> \./\1/' \ | |
| | sort \ | |
| | uniq -c \ | |
| | sort -nr \ |
| *.pyc | |
| mnist2500* | |
| build/ | |
| pip-log.txt | |
| text-documents/ | |
| asynchronous buffer forging | |
| anonymous identity injection | |
| asynchronous SQL skewing | |
| synchronous buffer analysis | |
| reverse jail fuzzing | |
| tainted state inspection | |
| multi-modal integrity recovery | |
| deep state engineering | |
| social state breaking | |
| monotonic state forging |
| #!/bin/bash | |
| sudo apt-get update | |
| sudo apt-get install -y byobu couchdb python-pip python-lxml | |
| sudo pip install -U tweepy couchdbkit restkit |
| """Evaluating the impact of PCA + whitening on low rank data""" | |
| import numpy as np | |
| from pprint import pprint | |
| from scikits.learn.datasets.samples_generator import make_regression_dataset | |
| from scikits.learn.pca import PCA | |
| from scikits.learn.linear_model import ElasticNetCV | |
| data_opts = { | |
| 'n_train_samples': 5000, |
| import numpy as np, scipy, scipy.sparse, numpy.linalg, scipy.optimize | |
| from scipy import weave | |
| def project_l1(lbda, sigma): | |
| "Project positive vector lbda to have l1 norm sigma" | |
| ll = -np.sort(-lbda) | |
| cs = 0. | |
| theta = 0 | |
| prevtheta = 0 |
| import time | |
| import sys | |
| import numpy as np | |
| from scipy import linalg | |
| from scikits.learn.linear_model import Lasso, lars_path | |
| from joblib import Parallel, delayed | |
| ################################################################################ | |
| # Utilities to spread load on CPUs |
| build | |
| *.so | |
| *.prof |
This is an experiment to highlight the dependency of the V-Measure value on the number of clusters of 2 independent uniform labelings for a finite number of samples.
Intuitively it seems that for finite number of samples the V-Measure is victim of some kind of birthday paradox that naive users might not be aware of.
Even if the maximum number of clusters considered (e.g. 10) is small with respect to the number of samples (e.g. 5000), the V-Measure of