Podcast Beta
Questions and Answers
What are the two key observations of the naive approach for the evaluation problem?
Only depends on the current state and only depends on the current state (Markov chain).
What is the main issue that arises when trying to compute the answer in the evaluation problem?
Combinatorial explosion due to the large number of state sequences to sum over.
How does the Forward Algorithm approach the evaluation problem?
By using Dynamic Programming and merging all states with the current state.
What does the Forward-Backward Algorithm aim to compute?
Signup and view all the answers
In the context of the decoding problem, how is the Viterbi Algorithm related to the evaluation problem?
Signup and view all the answers
What is the main drawback mentioned if one estimates all sequences in the decoding problem using Bayes' rule?
Signup and view all the answers
What is the purpose of the Baum-Welch Algorithm?
Signup and view all the answers
In the Baum-Welch Algorithm, what is done after initializing the guesses for the parameters?
Signup and view all the answers
What is the runtime complexity of the Viterbi algorithm?
Signup and view all the answers
What is the main idea behind Beam Search as an alternative to Viterbi?
Signup and view all the answers
What is an alternative to Beam Search for optimizing results on the Viterbi graph?
Signup and view all the answers
Why is Beam Search used in the context of Viterbi?
Signup and view all the answers
