I am trying to implement the following algorithm:
In a nutshell, we iterate over a set of 'states' to calculate a value for each.
Each state has multiple actions, and each multiple action has a 'transition probability' associated with it, such that if an action a is taken, there is a 0.8 chance it will proceed to a state s', and a 0.2 chance it will proceed to a different state.
When working through the algorithm, I am struggling to make a data structure that allows to locate a state, have a multiple actions available within that state, and for each action of that state, a probability matrix that when a particular action is taken to move to a new state, returns the probabilities of moving into a new state.
Any ideas to help me?
Many thanks.
Creating a Probability Transition Matrix
Page 1 of 12 Replies  859 Views  Last Post: 28 December 2012  10:15 PM
Replies To: Creating a Probability Transition Matrix
#2
Re: Creating a Probability Transition Matrix
Posted 28 December 2012  09:07 PM
A decision tree or a graph sounds like a good data structure to me. Perhaps a directed graph, specifically. You could also design a probability matrix using a 2D array or a HashMap.
#3
Re: Creating a Probability Transition Matrix
Posted 28 December 2012  10:15 PM
Thinking about this some more, you should look into Markov Chains with Eigenvalues and Eigenvectors. This is a topic in Linear Algebra.
Page 1 of 1
