HMMTeacher

HMMTeacher guides you in the learning of the mechanics of solving the main three HMM algorithms. From the input of the data of a pre-stated HMM, passing by choosing the questions asked and the algorithms to apply, to the final report, containing the step-by-step algorithms solution.

Step 1 - States
Hidden States
Step 2 - Matrix
Fill the matrix
Step 3 - Algorithm
Select an algorithm
Step 4 - Results
Final results

Select the alphabet of the observed sequence of states to use. Need more help?

You must select one of the 3 options to enable the inputs

You can change the data to try new things!

Either uppercase or lowercase letters are allowed

Is it your first time here? Ideally, use the manual to guide you in your first HMM model!

DNA Sequence

DNA alphabet

Alphabet of observed states: DNA Nucleotides (ACTG)

DNA observed sequence

Observed DNA sequence (One character per observed symbol. Max length: 100 characters.)

Hidden states:

Protein Sequence

Protein alphabet

Alphabet of observed states: One letter code for Proteins (ARNDBCEQZGHILKMFPSTWYV).

Protein observed sequence

Observed Protein sequence (One character per observed symbol. Max length: 100 characters.)

Hidden states:

Custom Sequence

Custom alphabet

Any custom alphabet, one character per symbol, without separators.
Example of alphabet of observed states for a coin, H (Heads) T (Tail): HT

Custom observed sequence

Observed Custom sequence (One character per observed symbol. Max length: 100 characters.)

Hidden states:

Input the data Need more help?

Any HMM is determined by three inputs, the vector of prior probabilities, the matrix of transition probabilities and the matrix of emission probabilities.

Do not leave any blank cells! All matrices must be complete with numbers. If you do not know the numbers, after filling all cells you know, leave the unknowns blank and press the Random button, to fill the blank cells with random numbers.

The trick to fill the matrices cells are: the sum of prior probabilities must sum one; The sum of each line in the transition (probabilities) matrix must sum one and; The sum of each line in the emission (probabilities) matrix must sum one.

You can use the check button to see if the summation is 1 or close to 1. When working with random float numbers in any programming language, rounding approximations may diverge the final expected result of 1.

Priors Vector

Transition matrix

Emission matrix

Select an algorithm Need more help?

Whenever you choose the backward algorithm, the forward algorithm is automatically chosen. This, because the backward algorithm need the result of the forward algorithm.

HMM can answer three main questions. Each question is answered by an algorithm. The questions and their algorithms are:

1. What is the probability of the observed sequence of elements, given the data? (Forward)

2. What is the probability that the observed state at position X was emitted by hidden state Y? (Backward) given the data.

3. What is the most probable sequence of hidden states given the data? (Viterbi) You can try more than one at a time.

You need to choose at least one algorithm.

HMM algorithms

Forward

Viterbi

Backward

Hidden State: Example: H0

Queried Position: Enter the hidden state. (ie 2)