What Is Hidden Markov Model (HMM)?

The hidden Markov Model (HMM) is a Markov model that's a bit more complicated than the average Markov. Do you know how there are only visible states in a standard Markov? Well, this one has hidden conditions that aren't very visible. They're hidden from you and also from everyone else. Why would you want to hide your state from everyone? Let's say you're trying to get into an exclusive club like Soho House or something. You might want to hide your membership status from other people because if they knew about it, they'd be more likely to treat you differently. The same goes for the HMM: when we train our models on data sets where the states are hidden (or unobserved), they can better predict whether or not someone belongs in an exclusive club—like Soho House! If you've ever been to a casino, you've probably seen a slot machine. In case you've never played one, it's pretty simple: you put your money in, pull the lever, and hope the device feels generous today. A hidden Markov model works much like a slot machine: it's based on probabilities and unpredictable results. The critical distinction between a Markov chain and the secret Markov model is that in the latter case, the state of a variable cannot be directly observed by an observer—it is "hidden," hence its name. Even though it's not visible to anyone else, it can still be inferred from the observable output. Hidden Markov models are like a puzzle. You have many pieces and have to figure out where they go. But instead of just one picture on the box, you have a bunch of photos—a whole pile of them—and each has missing parts. You can put some pieces together and see what happens, but then you might realize that there's a piece missing from the picture, and then you have to go back and look for it again. That's what it's like with HMMs: there's always something missing, but if you put in enough effort, you'll find it eventually!