#ANSQ2
data = np.load("data-2class.npz")
d = data['d']
l = data['l']

def sigma(a):
    return 1./(1.+np.exp(-a))

w = [0,-1,-1]

error = 0

for n in range(len(l)):
    update = l[n]*np.log(sigma(w[0]+w[1]*d[n,0]+w[2]*d[n,1]))+(1-l[n])*np.log(1-sigma(w[0]+w[1]*d[n,0]+w[2]*d[n,1]))
    error = error - update

print('The error is equal to:')
print(gradient)
#/ANSQ2