Pruning in simple terms is what you do to your rose plant in the garden cut off excess and those branches that do not lead to success.
2.15 imperfect real time decision
2.16 games that include an element of chance.
Tic Tack Toe: The element of chance depends on the other players ability. Foot Ball: Element of chance can be a pit in the ground (like marker points in Lawn Tennis), opponent getting cramps due to lack of salt in his/her body, rain favoring the chip player against the top spin player in tennis and so on and so forth.
3.0 UNIT III KNOWLEDGE REPRESENTATION 10
3.0 Introduction to Knowledge Representation
Knowledge representation centers around the AI research field of knowledge engineering and ontological reasoning. The machine or artifact cannot solve the problem without full and through knowledge of the world that it is expected to tackle. AI will be expected to represent objects, relations between them, types, happening s or events, the situation that causes the event, the cause and effect, time and states, meta data (data about data or knowledge about knowledge), upper and lower ontology.)
Definition of knowledge according to ontology:
As per ontology, knowledge is represented as the set of concepts and notions with in a domain and the relationship and correlations between the concepts under study.
Knowledge Representation (KR) research: (Pg 348, Ch 10, Part III)
The goal of KR Research is to use a set of symbols to represent a set of facts within a knowledge domain. For example, a mammal is represented as the animal with 4 legs and a neck within the knowledge domain of animal kingdom. A fish is represented as a life force that can swim in water with in the knowledge domain of aquatics. Now when we design a spotter agent, who is assigned to identify animals accurately on his Zoo trip, the knowledge representation plays a vital role to succeed.
Expressivity or the ability to express precisely is the key parameter in
choosing an existing KR ar creating a new one. The easier and compact way to express a fact or an element of knowledge within the semantics, syntax and lexical grammars of KR is to make it more expressive. But expressive languages need complex logic and algorithms for building the corresponding inferences. But once the KR becomes highly expressive, it becomes incomplete and inconsistent. Hence it is a good idea to keep a KR abstract or less expressive so that it would be complete and consistent. Example for highly expressive KR: Auto epistemic encompasses meaningful portions of knowledge made up of small, simple sequences of symbols called sentences. Autoepistemic is a formal logic for representing and reasoning meta knowledge (knowledge about knowledge). Autoepistemic can express knowledge about facts and also the lack of knowledge about facts if any. For example, in our case of Zoo animal spotter agent, to differentiate between a giraffe and a cheetah, mere sentences like it is a mammal, has 4 legs, has spots on skin, has a neck, has two eyes will represent lack of knowledge. This is because the agent needs uniquely identifying knowledge like has a long neck then can infer that as a giraffe or has sharp teeth, meat eater then can infer that as a cheetah.
Autoepistemic becomes stable model semantics in a simplified form, as it
gives semantics(meanings) to logic programming with positive as success and negation as failure.John McCarthy used advice-taker proposal logic as a declarative representation language. For problem solving, theorem-prover or model-generator was used.
Example for abstract, less expressive KR: Propositional logic consistent,
complete, well known for efficiently producing inferences with non complicated algorithms. Propositional logic can only express facts. We have already told you on page 7 that if you are trying to make the machines to think (part of meaning representation not knowledge representation) then, let us focus on ONTOLOGICAL ENGINEERING.