[Home]MarkovProperty

AIWiki | RecentChanges | Index | Preferences | Login

The Markov Property is true when the state of a system is sufficient for summarizing any effects that anything that happened in the past might have on the future. In other words, the system has no "memory" besides the state.

For example, if you are at your computer and you define the "state" of the computer to be your screen, you do not have the Markov Property. This is because in some circumstances (loading the file again), even if you know the entire "state", some additional knowledge of what happened in the past is helpful (what you saved) for predicting what will happen in the future.

If your computer was not connected to any network and it had infallible hardware, you could satisfy the Markov property by defining the state to include the entire memory plus everything on disk plus any changable information about the hardware that may not be stored in memory (such as the time). In this case, once you know the state, any additional information about what happened in the past will not help you predict what the system (the computer) will do.

Formally, the Markov property is "the future is independent of the past, given the state".


CategoryProbability


AIWiki | RecentChanges | Index | Preferences | Login
Edit text of this page | View other revisions
Last edited March 17, 2003 21:22 (diff)
Search: