Tanks a lot for your question.

Yes, you are right e = eps[127]. I actually did not set the maximum in the f array as threshold. argmax function gives the index of the maximum value. So, using argmax on f array I found out which index of probability is giving the maximum f value. The code for the last plot is here:

df1 = df[df['label']==0]

df2 = df[df['label']==1]

plt.figure()

plt.scatter(df1[0], df1[1], color='black', label = 'normal')

plt.scatter(df2[0], df2[1], color='red', label='anomalous')

plt.legend()

plt.show()