Hacker News new | past | comments | ask | show | jobs | submit login

The connection you are drawing is not incorrect in my opinion. But perhaps I can clarify.

The dynamics model in a KF is a linear system with an input and and output (often denoted u and y respectively).

A Hidden Markov Model typically does not have an input. There is an initial state, and then the system transitions "autonomously" (that's a term I'm borrowing from differential equations).

An "Markov model" with input is called a Markov Decision Process (MDP). And if the state of the system is not fully observable, then it's a POMDP.

So Kalman filters are most analogous to "belief updates" in POMDPs. The KF takes into account the known input to the system.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: