Topic: Hidden Markov Models

Why Baum-Welch Does Not Decrease Likelihood

Viterbi Paths Versus Pointwise Posterior Decoding

Viterbi Decoding as Joint MAP Inference

Stable HMM Computation with Scaling and Log Probabilities

Smoothing Hidden States with Forward-Backward Messages

Selecting the Number of Hidden States by Prediction

One Complete Baum-Welch Iteration by Hand

M-Step Updates for Categorical HMMs

Hidden Markov Models as Joint Probability Models

Handling Missing Observations in HMM Inference and Training