Floyd algorithm in Python3.5

#!/usr/bin/env python3
# Refer to: http://wiki.jikexueyuan.com/project/easy-learn-algorithm/floyd.html
# Comments: The Floyd algorithm can tell how much the nearest path between 2 vertex costs

N = 4           # number of total vertex
original = 1000 # init cost of unconnected vertex with an infinity

 # map in 2D grid to connect vertex with positive cost
cost = [[0, 2, 6, 4],
        [original, 0, 3, original],
        [7, original, 0, 1],
        [5, original, 12, 0]]
shortest_path = [[[1, 1], [1, 2], [1, 3], [1, 4]],
                 [[2, 1], [2, 2], [2, 3], [2, 4]],
                 [[3, 1], [3, 2], [3, 3], [3, 4]],
                 [[4, 1], [4, 2], [4, 3], [4, 4]]]

def remove_duplicates(val):
    output = []
    seen = set()
    for v in val:
        if v not in seen:
            output.append(v)
            seen.add(v)
    return output

for mid in range(N):
    for i in range(N):
        for j in range(N):
            if (cost[i][mid] + cost[mid][j]) < cost[i][j]:
                cost[i][j] = cost[i][mid] + cost[mid][j]
                shortest_path[i][j] = []
                shortest_path[i][j].extend(shortest_path[i][mid])
                shortest_path[i][j].extend(shortest_path[mid][j])

for i in range(N):
    #print(shortest_path[i])
    for j in range(N):
        shortest_path[i][j] = remove_duplicates(shortest_path[i][j])
    print(shortest_path[i])
    print(cost[i])