Command Line Interface

I have implemented a moderately clean and generalized command line interface for performing all necessary operations on the HMM.

Baum Welch Implemented

I have implemented the Baum Welch algorithm in its logarithmic form, and written unit tests which:

• confirm that the updated models are stochastic
• confirm the results agree with a hand-worked example found online

Now I can begin training models to my heart's desire.

Logarithmic Forward, Backward and Viterbi Algorithms

The basic functions for utilizing HMMs using logarithmic probabilities have been implemented, along with a LogHMM record type.

These functions all require a pre-constructed HMM, and have been tested on textbook examples.

Random Walk Sampler

I have implemented two random walk sample functions, sample-states and sample-emissions. Each take an HMM, and produce an infinite lazy seq of either states or emissions, by taking a random walk through the HMM, where each step is selected by a weighted RNG, based on the probabilities encoded in the HMM.

Forward, Backward and Viterbi Algorithms

The basic functions for utilizing HMMs have been implemented, along with an HMM record type.

These functions all require a pre-constructed HMM, and have been tested on textbook examples.