Unsupervised Neural Hidden Markov Models with a Continuous latent state space

TL;DR

This paper presents a neural approach to unsupervised Hidden Markov Models with continuous latent states, enhancing flexibility and interpretability while maintaining competitive performance on synthetic and real data.

Contribution

It introduces a novel neuralization method for continuous latent state HMMs, improving interpretability and flexibility over traditional models.

Findings

Comparable performance to LSTMs and GRUs

Easily interpretable model parameters

Effective on both synthetic and real data

Abstract

We introduce a new procedure to neuralize unsupervised Hidden Markov Models in the continuous case. This provides higher flexibility to solve problems with underlying latent variables. This approach is evaluated on both synthetic and real data. On top of generating likely model parameters with comparable performances to off-the-shelf neural architecture (LSTMs, GRUs,..), the obtained results are easily interpretable.