|
A policy is a plan of action for tackling political issues. It is often initated by a political party in government, which undergoes reforms and changes by interested actors (for example, opposition parties and lobby groups).
In AI planning and reinforcement learning, a policy prescribes a non-empty
deliberation (sequence of actions) given a non-empty sequence of states.
Types of policy include:
- stationary (resp. non-stationary)
- deterministic (resp. stochastic, randomized and sometimes non-deterministic)
- memoryless (e.g. non-stationary)
- causal (resp. non-causal)
- index
- opportunistic (resp. non-opportunistic)
These qualifiers can be combined, so for example you could have a stationary-memoryless-index policy.
|
