from random import choice

p_init = [-1,1,-1,1,-1,1,-1,1,-1] # pattern to learn

x = int(input("How many times do you want to train the network? "))

# initializes the weights' matrix with values set to 0
weights = list()
for n in range(len(p_init)):
    tmp = []
    for u in range(len(p_init)):
        tmp.append(0)
    weights.append(tmp)

# Hebbian Theory
def learn(pattern):
    i_index = -1
    for i in pattern:
        i_index += 1
        j_index = -1
        for j in pattern:
            j_index += 1
            if i_index != j_index:
                if i == j:
                    weights[i_index][j_index] += 1
                elif i != j:
                    weights[i_index][j_index] -= 1
                            
# Hopfield Update Rules
def update(pattern):
    i_index = -1
    for i in pattern:
        s = 0 # sum
        i_index += 1
        j_index = -1
        for j in pattern:
            j_index += 1
            if j != i:
                s += j * weights[i_index][j_index]
        if s >= 0:
            pattern[i_index] = 1
        elif s < 0:
            pattern[i_index] = -1



learn(p_init)

p = [1,1,-1,1,1,1,1,1,-1] # user pattern to be checked with the learned pattern (can be anything with the same length as p_init)
    
input_pattern = p.copy()

print("This is the pattern to be checked:", input_pattern)

for k in range(x): # trains for x time(s)
    if p != p_init:
        update(p)
    
if p == input_pattern:
    print("This may be a new pattern.")
else:
    print("I assume this pattern refers to the initial learned pattern.")