classify.py

#!/usr/bin/python3
'''
Created on Nov 1, 2016

 Generate data with two random gaussian samples with a tag.
 Basic code to test the tag recovery from the concatenated
 and shuffled samples.

@author: francis
'''

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

class Gauss(object):
    def __init__(self, mean=0, stddev=1):
        self.mean = mean
        self.stddev = stddev
    def gen_samples(self, n, tag=0):
        samples = self.stddev * np.random.randn(n) + self.mean;
        return pd.DataFrame({'tag': tag, 'samples': samples})
  

def main():
    ok = Gauss(200, 50).gen_samples(1000, 0)
    bad = Gauss(500, 50).gen_samples(100, 1)

    mix = pd.concat([ok, bad], ignore_index=True)
    
    mix = mix.reindex(np.random.permutation(mix.index))
    print(mix)
    
    plt.figure()
    plt.subplot(2, 1, 1)
    bins = 50 
    alpha = 0.5
    ok.samples.plot(kind='hist', alpha=alpha)
    bad.samples.plot(kind='hist', alpha=alpha)
    plt.subplot(2, 1, 2)
    mix.samples.plot(kind='hist', bins=bins)
    plt.savefig("mix.png")

if __name__ == '__main__':
    main()