053-How to pick optimal number of parameters.py

#!/usr/bin/env python
__author__ = "Sreenivas Bhattiprolu"
__license__ = "Feel free to copy, I appreciate if you acknowledge Python for Microscopists"

# https://www.youtube.com/watch?v=lGi07_thnqE

"""
How do we know how many clusters?
Use AIC/BIC
Bayesian information criterion (BIC) can be helpful to pick the right number of parameters.
BIC estimates the quality of a model using penalty terms for # parameters.
If we fit data with 100 gaussians we will be overfitting. 
BIC provides optimal vale if we do not know.
Looks asymptotic, pick the one at the elbow. 

"""
import numpy as np
import cv2

img = cv2.imread("images/Alloy.jpg")

img2 = img.reshape((-1,3))

from sklearn.mixture import GaussianMixture as GMM

#####################
#This part calculate BIC

n = 2
gmm_model = GMM(n, covariance_type='tied').fit(img2)

#print(gmm_model)

bic_values = gmm_model.bic(img2)
print(bic_values)



n_components = np.arange(1,10)

gmm_models = [GMM(n, covariance_type='tied').fit(img2) for n in n_components]

#print(gmm_models)


print(GMM.bic(img2))

from matplotlib import pyplot as plt
plt.plot(n_components, [m.bic(img2) for m in gmm_models], label='BIC')
plt.xlable('n_components')