Uncertainty

Instinctive:

all major components and their relations are determined when the system is formed, and remain unchanged afterwards

Intelligent:

all major components and their relations are adaptive to the environment. The system learns, under the assumption that in general the future will be similar to the past.

Intelligence is the ability to adapt to the environment with insufficient knowledge and resources.

• The Representation Language

• Imperfect Observation of the World

• Ignorance, Laziness, Efficiency

• provided by the designer of the agent

Sources of uncertainty:

• the are more states in the real world than can be represented by the language -> a state described in the representation language referes to multiple states in the real world

Observation can be:

• partial: sensor does not see 360 degree

• ambigious: sensor output has multiple possible interpretations

• incorrect

• designer abstracts complexity from the real world

• might lead to incorrect results

1. non-deterministic model

   • set of possible values

2. probabilistic model

   • probabilistic distribution over set of possible values

Probability: The theory of probability provides a way of summarizing the uncertainty that comes from our laziness and ignorance

Utility: Every state has a degree of usefulness to an agent

Decision Theorie = Probability Theory + Utility Theory

An agent is rational if and only if it chooses the action that yields the highest expected utility, averaged over all the possible outcomes of the action.

• A proposition (patient has cavity) together with a probabilistic value (p)

• probabilistic distribution over all the possible states

• dentist with three propositions: cavity, toothache, catch

• dentists belief state consist of 8 possible states

  • each has probability

• given n propositions -> 2^n states in belief state

• many probabilites

Better:

• explicitly represent independence among propositions to reduce the number of probabilities defining a belief state -> Bayesian networks

• Notice that Cavity is the “cause” of both Toothache and Catch, and represent the causality links explicitly

• Arrow means: Cavity has influences probability of Toothcache and Catch

  • given the fact that the patient suffers from cavity, the patient has toothcache with 60% and catch with 90%

  • toothcache and catch are independent given cavity

• 5 probability values instead of 7

https://de.wikipedia.org/wiki/Bayessches_Netz

Yes, because we can compute the full joint distribution: