def __run_bayesfmm(data, iterations, save_diagnostics, output_dir, burnin, km2):
gaussian_fmm = fmm.GaussianFiniteMixtureModel()
pi, theta, sigma = gaussian_fmm.run(data, k=2, iterations=iterations)
outfilename = None
if save_diagnostics:
outfilename = '{0}/diagnostics_{1}_{2}.png'.format(output_dir, '{0}', '{1}')
#generate diagnostics
diagnostics.stationarity_plot(pi, ylab=r'\pi',
filename=(outfilename.format('stationarity','pi')
if outfilename is not None else None))
diagnostics.stationarity_plot(theta, ylab=r'\theta',
filename=(outfilename.format('stationarity','theta')
if outfilename is not None else None))
diagnostics.stationarity_plot(sigma, ylab=r'\Sigma',
filename=(outfilename.format('stationarity','sigma')
if outfilename is not None else None))
#determine model parameters
pi_mean, theta_mean, sigma_mean = fmm.calculate_posterior_means(pi, theta, sigma, burnin)
# print pi_mean, theta_mean, sigma_mean
em.plot_against_kmeans(data, theta, km2,
filename='{0}/faithful_bayesfmm_kmeans.png'.format(output_dir))
fmm.draw_contour_plots(pi_mean, theta_mean, sigma_mean,
filename='{0}/faithful_bayesfmm_model.png'.format(output_dir))
def __run_em(data, output_dir, km2):
theta, sigma, pi = em.gaussian_em_2(data, max_reps=100)
em.plot_against_kmeans(
data,
theta,
km2,
filename='{0}/faithful_em_kmeans.png'.format(output_dir))
em.draw_contour_plots(
theta[-1],
sigma[-1],
pi[-1],
filename='{0}/faithful_em_model.png'.format(output_dir))
def __run_bayesfmm(data, iterations, save_diagnostics, output_dir, burnin,
km2):
gaussian_fmm = fmm.GaussianFiniteMixtureModel()
pi, theta, sigma = gaussian_fmm.run(data, k=2, iterations=iterations)
outfilename = None
if save_diagnostics:
outfilename = '{0}/diagnostics_{1}_{2}.png'.format(
output_dir, '{0}', '{1}')
#generate diagnostics
diagnostics.stationarity_plot(
pi,
ylab=r'\pi',
filename=(outfilename.format('stationarity', 'pi')
if outfilename is not None else None))
diagnostics.stationarity_plot(
theta,
ylab=r'\theta',
filename=(outfilename.format('stationarity', 'theta')
if outfilename is not None else None))
diagnostics.stationarity_plot(
sigma,
ylab=r'\Sigma',
filename=(outfilename.format('stationarity', 'sigma')
if outfilename is not None else None))
#determine model parameters
pi_mean, theta_mean, sigma_mean = fmm.calculate_posterior_means(
pi, theta, sigma, burnin)
# print pi_mean, theta_mean, sigma_mean
em.plot_against_kmeans(
data,
theta,
km2,
filename='{0}/faithful_bayesfmm_kmeans.png'.format(output_dir))
fmm.draw_contour_plots(
pi_mean,
theta_mean,
sigma_mean,
filename='{0}/faithful_bayesfmm_model.png'.format(output_dir))
def __run_em(data, output_dir, km2):
theta, sigma, pi = em.gaussian_em_2(data, max_reps=100)
em.plot_against_kmeans(data, theta, km2,
filename='{0}/faithful_em_kmeans.png'.format(output_dir))
em.draw_contour_plots(theta[-1], sigma[-1], pi[-1],
filename='{0}/faithful_em_model.png'.format(output_dir))
