from math import exp, log
from random import random

def sigmoid(x):
	return 1/(1+exp(-x))

def predict(coef, x):
	y = 0
	for i in range(len(coef)):
		y = y + coef[i]*x[i]
	return y

def update_coef(coef, x, y, learningRate):
	yguess = []
	for i in range(len(x)):
		yguess.append(sigmoid(predict(coef, x[i])))
	for i in range(len(coef)):
		gradient = 0
		for j in range(len(x)):
			eroare = yguess[i] - y[i]
			gradient = gradient + (eroare * x[j][i])
		coef[i] = coef[i] - (learningRate * gradient)
	return coef

def cost(x, y, coef):
	c = 0
	for i in range(len(x)):
		yguess = sigmoid(predict(coef, x[i]))
		t1 = y[i] - log(yguess)
		t2 = (1 - y[i]) * log(1-yguess)
		c = c - t1 - t2
	return c / len(x)

def label(x, coef):
	if sigmoid(predict(coef, x)) < 0.5:
		return 0
	return 1

def train(x, y, learningRate, numarEpoci, eroareMinima):
	coef = [random() for i in range(len(x[0]))]
	costCurent = cost(x, y, coef)
	while costCurent > eroareMinima and numarEpoci > 0:
		update_coef(coef, x, y, learningRate)
		costCurent = cost(x, y, coef)
		numarEpoci = numarEpoci - 1
	return coef

def accuracy(x, y, coef):
	suma = 0
	for i in range(len(x)):
		yguess = label(x[i], coef)
		if yguess == y[i]:
			suma = suma + 1
	return suma / len(x)

if __name__ == '__main__':
	#x = [[-1.11,-0.77], [-0.76, -1.25], [-0.06, 0.19], [0.62, 1.15], [1.32, 0.67]]
	x = [[1.0, 0.00662, 3.38e-05], [1.0, 0.0081, 6.375e-05], [1.0, 0.00273, 1.57e-05], [1.0, 0.00382, 3.005e-05], [1.0, 0.00501, 3.648e-05], [1.0, 0.00453, 3.481e-05], [1.0, 0.00526, 6.016e-05], [1.0, 0.0087, 4.743e-05], [1.0, 0.0124, 0.00011434], [1.0, 0.00476, 4.216e-05]]
	#y = [0,0,1,1,1]
	y = [28.199, 11.218, 23.856, 11.27, 31.0, 27.729, 16.816, 18.169, 17.0, 12.0]
	coef = train(x, y, 0.0005, 100000, 0.5)
	#yout = label([-0.41, -0.28], coef)
	yout = sigmoid(predict(coef, [[1.0, 0.00355, 2.694e-05], [1.0, 0.00644, 4.285e-05], [1.0, 0.00495, 2.524e-05]]))
	print(yout)