I have been learning about Search Problems, and have been struggling with the difference between the two.
Search Problems invloving uncertainties can be modeled using MDPs, and it doesn't matter if we get a finite search tree or an infinite search tree, Value Iteration and Policy Iteration algorithms can be used on them. Now I am confused in the part as to how Expectimax algorithm fit into all this? I know the difference between them mathematically as in discount factor in case of Expectimax algorithm is 1, whereas it can be anything in the case of Value Iteration. There are only terminal utilities in case of Expectimax Algorithm, however, in Value Iteration we have reward associated with each transition.
I have tried discussing it with my fellow peers and they told me that "Value Iteration algorithm is the generalization of Expectimax algorithm."
What I am not able to understand is that what is the significance of the Expectimax Algorithm? What is a real world problem in which Expectimax Algorithm will work but Value Iteration will not? To be precise in the set of Search Problem with uncertainities, what is the Venn diagram of MDPs, Value Iteration and Expectimax algorithm?
