Gibbs Sweep (Seeded)

Problem

One Gibbs sweep for a two-variable binary model with joint table joint, where joint[x][y] is the joint probability. Starting from x = start (0 or 1), sample y from P(Y | X = x), then resample x from P(X | Y = y) using random.Random(seed). Return the new state as [x, y].

Hint

Normalize each conditional by the appropriate row or column sum.

Test cases

case 1

in: [[[0.4,0.1],[0.2,0.3]],0,42]

expected: [1,0]

case 2

in: [[[0.4,0.1],[0.2,0.3]],1,7]

expected: [1,1]

case 3

in: [[[0.4,0.1],[0.2,0.3]],0,0]

expected: [0,0]

case 4

in: [[[0.4,0.1],[0.2,0.3]],1,123]

expected: [1,1]

Starter codePython

import random


def gibbs_sweep(joint, start, seed):
    # Your code here
    pass

4 test cases · run your code against them in the editor