I have a transition matrix of 40 states that gives transition probability to move from one state to another. How can we compute that after n time-steps the system will be in what state.
Asked
Active
Viewed 98 times
1 Answers
0
Raise the matrix to the power of 40 and then multiply that with the initial state vector.
So, if v0 is the initial state vector, M is the trasition matrix, and v40 is the vector after 40 steps then:
M^40 * v0 = v40

dimm
- 1,792
- 11
- 15
-
I am trying, but in my case it does not seem to give any good results. I have created a transition matrix using a time series of 8820 steps. And i am predicting 8821 step as v0\* M^8821 and then using choosing maximum out of v40. v0 is taken as a vector of all zeros with only starting state set to 1. Am I doing it right? Because for all values ahead of 8821 and so on...it is predicting the same wrong state. – user6460588 Aug 21 '16 at 09:52