Download Class Notes on Predictive Logic and System Specifications (October 8, 2003) and more Study notes Computer Science in PDF only on Docsity! Class Notes for 8th of October, 2003 Prepared by Kahkashan Shaukat For good programming practice: provide specifications as in software engineering The specifications are: 1. Input: what kind of information or knowledge is provided by the system 2. Queries: what kind of questions may be asked 3. Define what it means for something to be a correct answer to a query 4. Level of confidence that the answer to the query is correct – that our implementation is correct It is very important to write rules in a systematic manner otherwise on rule may change the effect of a lot of rules as they are non-monotonic. For the Prediction/Projection System: 1. Input: There are two types of inputs a) Domain Description Language: Denoted by D. These consist of a causes f if p1,…..pm, qq1……qr here a is an action e.g. shoot, load, pickup something p1,…..pm, qq1……qr q are fluents and f is a fluent literal fluents are properties of the world, that is those that change the world, e.g. alive, loaded,on(A|B) etc. a fluent or its negation is called a fluent literal e.g. alive, alive etc. If we have: i) a causes f a causes g it means a causes f,g f and g here are conjunctive and not disjunctive ii) a causes f if p a causes g if q - then p and q are disjunctive b) Initial State Representation: Denoted by O. Initially f, f is a fluent literal 2. Queries: f after a1….an. implies that after performing actions a1….an will f be true. eg. Will the turkey be alive after load, shoot? 3. Correctness: (D,O) |= Q that is (D,O) entails Q. When (D,O) |= Q is true or false? (D,O) are knowledge then when can you conclude Q? So, is it possible to conclude Q given D,)? So, (D,O) |= Q if and only if 1(D,O) |=smodels tr(Q). so whenever, we have (D,O) |= f we can prove (D,O) |= tr(Q) Interpretation of Initially: For example: a causes f if p a causes f if p and the query is: will f be true after a? This query does not find any information about p or p. So, if the initial information is incomplete it will be impossible to tell about the consequences. Because initial information give information about the initial states and if it is incomplete that means multiple states may be possible. Define initial state and state transitions then it is easier to define the problem. A state is a set of fluents; whatever exists there is true, otherwise it is false.