from multiprocessing import Process, Manager

import numpy as np

from scipy.stats import logistic

import argparse

import time

# gradient descent to update theta

def gradientDescent(parameters, result):

result[parameter] = learning_rate * summition / number_of_examples

# logistic function

def sigmoid(value):

return logistic.cdf(value)

def hypothesis(X, theta):

hx = np.dot(theta, X)

return sigmoid(hx)

#Initialize argument parser

parser = argparse.ArgumentParser()

parser.add_argument("-f", "--filename", help="input file name")

parser.add_argument("-l", "--learning_rate", type=float, help="learning rate")

parser.add_argument("-i", "--iterations", type=int, help="number of iteration")

args = parser.parse_args()

if args.filename:

filename = args.filename

else:

print('Please specify a file name')

print('Execute "python logistic.py -h" for help')

exit()

if args.learning_rate:

learning_rate = args.learning_rate

else:

learning_rate = 1

print('Using default learning rate', learning_rate)

if args.iterations:

iterations = args.iterations

else:

iterations = 100

print('Using default iterations', iterations)

# read data file

data_array = np.loadtxt(filename, delimiter=',')

# number of training examples and features

number_of_examples = data_array.shape[0]

number_of_parameters = data_array.shape[1]

# add column of ones to dataframe make looping symmetric

ones = np.ones(number_of_examples)

data_array = np.insert(data_array, 0, values=ones, axis=1)

# initialize theta to zeroes array

theta = np.zeros(number_of_parameters)

parameters_split = np.array_split(list(range(len(theta))), 4)

if __name__ == '__main__':

manager = Manager()

results = manager.dict()

start = time.time()

for i in range(iterations):

processes = []

# multiple process to caculate theta (4 sublist of theta)

for j in range(len(parameters_split)):

processes.append(Process(target=gradientDescent, args=(parameters_split[j], results)))

theta_diff = []

for p in processes:

p.start()

for p in processes:

p.join()

# here to merge

for j in range(len(theta)):

theta_diff.append(results[j])

theta = theta - theta_diff

print("round: ", i)

end = time.time()

#print total time

print(end - start)