genimages.py

import numpy as np
import matplotlib.pyplot as plt

N=400  # number of data points - you can increase this if you want to
       # learn better features (but it will take longer).
D=16   # dimensionality of the data

np.random.seed(0)

# Define the basic shapes of the features

m1 = [0, 0, 1, 0,
      0, 1, 1, 1,
      0, 0, 1, 0,
      0, 0, 0, 0]

m2 = [0, 1, 0, 0,
      0, 1, 0, 0,
      0, 1, 0, 0,
      0, 1, 0, 0]

m3 = [1, 1, 1, 1,
      0, 0, 0, 0,
      0, 0, 0, 0,
      0, 0, 0, 0]

m4 = [1, 0, 0, 0,
      0, 1, 0, 0,
      0, 0, 1, 0,
      0, 0, 0, 1] 

m5 = [0, 0, 0, 0,
      0, 0, 0, 0,
      1, 1, 0, 0,
      1, 1, 0, 0] 

m6 = [1, 1, 1, 1,
      1, 0, 0, 1,
      1, 0, 0, 1,
      1, 1, 1, 1] 

m7 = [0, 0, 0, 0,
      0, 1, 1, 0,
      0, 1, 1, 0,
      0, 0, 0, 0]

m8 = [0, 0, 0, 1,
      0, 0, 0, 1,
      0, 0, 0, 1,
      0, 0, 0, 1]

nfeat = 8 # number of features
rr = 0.5 + np.random.rand(nfeat, 1) * 0.5 # weight of each feature between 0.5 and 1
mut = np.array([rr[0] * m1, rr[1] * m2, rr[2] * m3, rr[3] * m4, rr[4] * m5,
                rr[5] * m6, rr[6] * m7, rr[7] * m8])
s = np.random.rand(N, nfeat) < 0.3 # each feature occurs with prob 0.3 independently 

# Generate Data - The Data is stored in Y

Y = np.dot(s, mut) + np.random.randn(N, D) * 0.1 # some Gaussian noise is added 

nrows = 13
for k in range(16):
    plt.subplot(4, 4, k + 1)
    plt.imshow(np.reshape(Y[k], (4, 4)), cmap=plt.gray(), interpolation='none')
    plt.axis('off')

plt.show()