Groups
Category
A Hidden Markov Model (HMM) describes sequences where you cannot see the true state directly, but you can observe outputs generated by those hidden states.
CTC loss trains sequence models when you do not know the alignment between inputs (frames) and outputs (labels).