test_hmm.py

import json
import os
import sys

from myhmm import MyHmm

models_dir = os.path.join('.', 'models')  #

seq0 = ('Heads', 'Heads', 'Heads')
seq1 = ('Heads', 'Heads', 'Tails')
seq2 = ('Heads', 'Tails', 'Heads')
seq3 = ('Heads', 'Tails', 'Tails')
seq4 = ('Tails', 'Heads', 'Heads')
seq5 = ('Tails', 'Heads', 'Tails')
seq6 = ('Tails', 'Tails', 'Heads')
seq7 = ('Tails', 'Tails', 'Tails')

observation_list = [seq0, seq1, seq2, seq3, seq4, seq5, seq6, seq7]

if __name__ == '__main__':
    # test the forward algorithm and backward algorithm for same observations and verify they produce same output
    # we are computing P(O|model) using these 2 algorithms.
    model_file = "coins.json"  # this is the model file name - you can create one yourself and set it in this variable
    hmm = MyHmm(os.path.join(models_dir, model_file))
    print("Using the model from file: ", model_file,
          " - You can modify the parameters A, B and pi in this file to build different HMM models")

    total1 = total2 = 0  # to keep track of total probability of distribution which should sum to 1
    for obs in observation_list:
        p1 = hmm.forward(obs)
        p2 = hmm.backward(obs)
        total1 += p1
        total2 += p2
        print("Observations = ", obs, " Fwd Prob = ", p1, " Bwd Prob = ", p2,
              " total_1 = ", total1, " total_2 = ", total2)

    # test the Viterbi algorithm
    observations = seq6 + seq0 + seq7 + seq1  # you can set this variable to any arbitrary length of observations
    prob, hidden_states = hmm.viterbi(observations)
    print("Max Probability = ", prob, " Hidden State Sequence = ", hidden_states)

    print("Learning the model through Forward-Backward Algorithm for the observations", observations)
    model_file = "random.json"
    hmm = MyHmm(os.path.join(models_dir, model_file))
    print("Using the model from file: ", model_file,
          " - You can modify the parameters A, B and pi in this file to build different HMM models")
    hmm.forward_backward(observations)

    print("The new model parameters after 1 iteration are: ")
    print("A = ", hmm.A)
    print("B = ", hmm.B)
    print("pi = ", hmm.pi)