Show Sidebar Hide Sidebar

# Gaussian Mixture Model Selection in Scikit-learn

This example shows that model selection can be performed with Gaussian Mixture Models using information-theoretic criteria (BIC). Model selection concerns both the covariance type and the number of components in the model. In that case, AIC also provides the right result (not shown to save time), but BIC is better suited if the problem is to identify the right model. Unlike Bayesian procedures, such inferences are prior-free.

In that case, the model with 2 components and full covariance (which corresponds to the true generative model) is selected.

#### New to Plotly?¶

You can set up Plotly to work in online or offline mode, or in jupyter notebooks.
We also have a quick-reference cheatsheet (new!) to help you get started!

### Version¶

In [1]:
import sklearn
sklearn.__version__

Out[1]:
'0.18.1'

### Imports¶

In [2]:
import plotly.plotly as py
import plotly.graph_objs as go
from plotly import tools

import numpy as np
import itertools
import math

from scipy import linalg
from sklearn import mixture


### Calculations¶

In [3]:
# Number of samples per component
n_samples = 500

# Generate random sample, two components
np.random.seed(0)
C = np.array([[0., -0.1], [1.7, .4]])
X = np.r_[np.dot(np.random.randn(n_samples, 2), C),
.7 * np.random.randn(n_samples, 2) + np.array([-6, 3])]

lowest_bic = np.infty
bic = []
n_components_range = range(1, 7)
cv_types = ['spherical', 'tied', 'diag', 'full']
for cv_type in cv_types:
for n_components in n_components_range:
# Fit a Gaussian mixture with EM
gmm = mixture.GaussianMixture(n_components=n_components,
covariance_type=cv_type)
gmm.fit(X)
bic.append(gmm.bic(X))
if bic[-1] < lowest_bic:
lowest_bic = bic[-1]
best_gmm = gmm

bic = np.array(bic)
color_iter = itertools.cycle(['navy', 'turquoise', 'cornflowerblue',
'darkorange'])
clf = best_gmm


### Plot Results¶

In [4]:
fig = tools.make_subplots(rows=2, cols=1,
print_grid=False,
subplot_titles=('BIC score per model',
'Selected GMM: full model, 2 components'))


Plot the BIC scores

In [5]:
for i, (cv_type, color) in enumerate(zip(cv_types, color_iter)):
xpos = np.array(n_components_range) + .2 * (i - 2)
trace = go.Bar(x=xpos, y=bic[i * len(n_components_range):
(i + 1) * len(n_components_range)],
marker=dict(color=color, line=dict(color='black' , width=1)),
name=cv_type)
fig.append_trace(trace, 1, 1)

fig['layout']['yaxis1'].update(range=[bic.min() * 1.01 - .01 * bic.max(), bic.max()],
zeroline=False, showgrid=False)

xpos = np.mod(bic.argmin(), len(n_components_range)) + .65 +\
.2 * np.floor(bic.argmin() / len(n_components_range))

fig['layout'].update(annotations=[dict(x=xpos, y=bic.min(),
text='*', yref='yaxis1', xref='xaxis1')],
hovermode='closest', height=800)


Plot the winner

In [6]:
Y_ = clf.predict(X)

for i, (mean, cov, color) in enumerate(zip(clf.means_, clf.covariances_,
color_iter)):
v, w = linalg.eigh(cov)
if not np.any(Y_ == i):
continue
trace = go.Scatter(x=X[Y_ == i, 0], y=X[Y_ == i, 1],
mode='markers',
showlegend=False,
marker=dict(color=color,
line=dict(color='black' , width=1)))
fig.append_trace(trace, 2, 1)

# Plot an ellipse to show the Gaussian component
v = 2. * np.sqrt(2.) * np.sqrt(v)
a =  v[1]
b =  v[0]
x_origin = mean[0]
y_origin = mean[1]
x_ = [ ]
y_ = [ ]

for t in range(0,361,10):
x_.append(x)
y_.append(y)

elle = go.Scatter(x=x_ , y=y_, mode='lines',
showlegend=False,
line=dict(color=color, width=2))
fig.append_trace(elle, 2, 1)

fig['layout']['yaxis2'].update(zeroline=False, showgrid=False)
fig['layout']['xaxis2'].update(zeroline=False, showgrid=False)

In [7]:
py.iplot(fig)

Out[7]:
Still need help?