3

I fitted a 2-states-HMM model last week, and generate a bunch of 1s and 0s, but I forgot to store its parameters (transition matrix). Now, I only got these 1s and 0s, how do I backward/reverse-engineering to estimates these transition matrix?

Things that I have:

  1. the input data
  2. the outputed 0s and 1s from a fitted 2-states-HMM

What I want:

  1. The transition matrix of the two-states HMM model.

Please see the exact input and output data here:

https://colab.research.google.com/drive/1N-PKaeHMVI4S1VU7fgcgDqRT5L3h8R0L?usp=sharing

And go to the bottom which it says Input_data and Output_data

kou
  • 159
  • Is the transition matrix graph kown? – Cesareo Sep 11 '22 at 23:29
  • I am trying to figure out the transition matrix of the two states HMM model – kou Sep 11 '22 at 23:32
  • Is the output a generated ongoing sequence? Or is it restarting after a certain length? (Prior may be relevant then) – DrShredz Sep 11 '22 at 23:37
  • A single on-going sequence. – kou Sep 12 '22 at 00:29
  • Can you make the input-output data available? – Cesareo Sep 12 '22 at 10:57
  • Yes, please see this link: https://colab.research.google.com/drive/1N-PKaeHMVI4S1VU7fgcgDqRT5L3h8R0L?usp=sharing – kou Sep 12 '22 at 15:01
  • And go to the bottom which it says Input_data and Output_data – kou Sep 12 '22 at 15:02
  • could you remind me, how exactly does input data come into a picture? So you have a Markov Chain with two states, and you have a noisy observation of a state, right? – SBF Sep 12 '22 at 16:15
  • Good question, I have the input data which is a numerical sequence, and I am trying to fit a HMM onto this numerical sequence, and I am trying to determine whether the sequence is of state 1 or state 0 – kou Sep 12 '22 at 17:16
  • I guess then you don't need the input data to get the lost HMM. You can just generate data using that HMM and fit to it – SBF Sep 14 '22 at 10:17

0 Answers0