Hidden Markov Model Homework: State Space S={1,2,3}, Alphabet A={a,b,c}

  • Thread starter Thread starter daneault23
  • Start date Start date
  • Tags Tags
    Models
Click For Summary
SUMMARY

The discussion focuses on defining a Hidden Markov Model (HMM) with a state space S={1,2,3} and an alphabet A={a,b,c}. The transition matrix P is specified as P = [[0.5, 0.5, 0], [0, 1, 0], [0, 0, 1]], and the initial probability vector ∏ is [1, 0, 0]. The probabilities for observations are given as b1(a) = 1/2, b1(b) = 1/2, b1(c) = 0, b2(a) = 1/2, b2(b) = 0, b2(c) = 1/2, b3(a) = 0, b3(b) = 1/2, b3(c) = 1/2. The probability of observing the sequence O = (a, b, c) is calculated to be 1/16, primarily derived from the state sequence Q = (1, 3, 2).

PREREQUISITES
  • Understanding of Hidden Markov Models (HMM)
  • Familiarity with transition matrices and initial probability vectors
  • Knowledge of emission probabilities in HMMs
  • Ability to calculate probabilities using HMM parameters
NEXT STEPS
  • Study the Viterbi algorithm for finding the most likely state sequence in HMMs
  • Learn about the Forward algorithm for calculating the probability of an observation sequence
  • Explore the Baum-Welch algorithm for training HMMs
  • Investigate applications of HMMs in natural language processing and bioinformatics
USEFUL FOR

Students and professionals in data science, machine learning practitioners, and anyone interested in understanding or applying Hidden Markov Models in various fields such as speech recognition and sequence analysis.

daneault23
Messages
32
Reaction score
0

Homework Statement


Define a Hidden Markov Model with the following parameters: State space S={1,2,3}, alphabet A = {a,b,c,}, P = 0 .5 .5
1 0 0
0 1 0

initial probability vector, ∏ = 1 0 0

b1(a) = 1/2 ; b1(b) = 1/2 ; b1(c) = 0
b2(a) = 1/2 ; b2(b) = 0; b2(c) = 1/2
b3(a) = 0; b3(b) = 1/2 ; b3(c) = 1/2

List all possible state sequences Q = (q1; q2; q3). Consider the observations
O = (a; b; c). What is the probability to observe this sequence in the model, i.e.
find P(O|λ)? Given these observations, which state sequence has most likely
generated it, i.e., find P(Q|O)?

Homework Equations



Remember the collection of parameters of a Hidden Markov Model is denoted λ = (P, B, ∏)

The Attempt at a Solution



Since ∏ has a 1 in the first entry, we always start in state 1. Thus, the probabilities of starting in other states, (P(O=abc|Q=312)*P(312) for example = 0), so we only need to find P(O=abc|Q=123)*P(Q=123) + P(O=abc|Q=132)*P(132). Well, right off the bat, the probability of going from state 2 to observation b is 0 so P(O=abc|Q=123)*P(Q=123)=0 already. Therefore, the only thing left is P(O=abc|Q=132)*P(132)=.5^4=1/16, so the probability to observe the sequence (a,b,c) in this model is 1/16.

Am I doing something wrong here?
 
Physics news on Phys.org
I just realized that the sequence could also be given by Q=111, 112, 113, 121, 122, and 133. It wasn't as easy as I originally thought.
 

Similar threads

  • · Replies 12 ·
Replies
12
Views
2K
  • · Replies 3 ·
Replies
3
Views
1K
Replies
0
Views
2K
  • · Replies 11 ·
Replies
11
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 4 ·
Replies
4
Views
2K
  • · Replies 4 ·
Replies
4
Views
2K
  • · Replies 4 ·
Replies
4
Views
3K
Replies
5
Views
4K
  • · Replies 80 ·
3
Replies
80
Views
8K