# Can the hidden states of a HMM be interpreted as number of clusters underlying the data?

Trying to understand the physical significance of the number of hidden states of a HMM. Should they be interpreted as number of clusters in the data? If not, why? Or they should be interpreted as the number of components of a GMM?

Best

• Depending on your application, the number of states in HMM can indeed be the number of clusters in your data. Re: "Or they should be interpreted as the number of components of a GMM?" -- GMM and HMM are two different models, it is important to distinguish between the two. – kedarps Aug 15 '18 at 15:43