WorksheetsFlipped Classroom activity
Total questions: 14
Worksheet time: 7mins
What is a Hidden Markov Model (HMM) primarily used for?
Predicting stock market trends
Identifying hidden states in sequential data
Analyzing image pixels
Calculating genetic mutations
In an HMM, what do the emission probabilities represent?
The probability of transitioning from one state to another
The likelihood of observing a certain output given a state
The initial probabilities of each state
The total probability of all states
What algorithm is used in HMMs to find the most likely sequence of hidden states given a sequence of observations?
Baum-Welch Algorithm
Viterbi Algorithm
Forward Algorithm
Backward Algorithm
Which of the following is NOT a component of an HMM?
Transition matrix
Emission matrix
Initial state distribution
Observation probability matrix
In a first-order HMM, the probability of being in a particular state depends only on the previous state.
In the context of HMMs, the term “state space” refers to the set of all possible observations.
In which type of HMM are observations modeled as continuous variables, typically using a Gaussian distribution?
Discrete HMM
Continuous HMM
Gaussian HMM
Hidden Markov Model with Multinomial Emissions
In a Left-to-Right HMM, how are state transitions constrained?
Transitions can occur only from the current state to previous states.
Transitions can occur only from the current state to future states.
Transitions can occur only from the current state to any other state.
Transitions are not allowed between states.
Which of the following statements regarding Markov Models is/are True?
Transition probabilities specify how the state evolves over time.
Transition probabilities are the same at all times
Markov models have a choice over actions to take.
In Markov models, the future and past states are dependent.
Which of the following statements is/are True?
Markov chains are not practically relevant for most agents since full observations are needed to update beliefs.
Hidden Markov Models have an underlying Markov chain over observations
You can observe outputs at each time step in Hidden Markov Models
Hidden Markov Models have an underlying Markov chain over states.
An HMM is defined by:
Initial distribution, emissions
Initial distribution, transitions
Initial distribution, transitions, emissions
Initial distribution, transitions, emissions, actions
Which of the following statements regarding HMMs is/are True?
Future state is dependent on the past state given the present state.
Evidence variables are guaranteed to be independent
Current observation is independent of all else given the current state.
Current observation is dependent on past states.
Which of the following statements is/are True?
Filtering, or monitoring, is the task of tracking the distribution over time.
As we get observations and beliefs get reweighted, uncertainty in the hidden state "increases"
The initial belief state is usually set to be uniform
Filtering takes all past observations into account to update belief of hidden state.
Which of the following statements is/are True regarding online belief updates?
Every time step, we start with current P(X|evidence)
We update for time and evidence with normalization
The forward algorithm updates for time and evidence with normalization
The forward algorithm does both time and evidence at once
