Dept Computer Science Korea Univ Intelligent Information System

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. A I (Artificial Intelligence) Professor I. J. Chung 9/8/2021 1

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. AI (Artificial Intelligence) Symbolic Reasoning under uncertainty A, B, C : event of Rich text Problems posed by uncertainty, fuzzy & changing knowledge Conventional reasoning system : 1 st order predicate logic, MR NMR Inference with incomplete knowledge(p. 195 Rich) To solve problems with uncertainty & fuzziness Knowledge base may not be grow monotonically as new assertions are made Qualititive reasoning rather than quantitive reasoning 9/8/2021 Artificial Intelligence 2

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. AI (Artificial Intelligence) MR Set of axioms & infer their consequences Addition of new axioms → true statements ↑ Not adequate to model the fuzzy, uncertain data based on assumptions, beliefs, impreciseness, etc. Some beliefs may change as time passes by 9/8/2021 Artificial Intelligence 3

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. AI (Artificial Intelligence) Default reasoning Draw conclusions based on what is mostly likely to be : DR NML Abduction Inheritance DL NML 1 st order logic + modal operator M (“is consistent with”) Eg. ∀x, y : Related(x, y) ∧M getalong(x, y) → wildefend(x, y) 9/8/2021 Artificial Intelligence 4

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. AI (Artificial Intelligence) DL A: B C If A is provable & it is consistent to assume B, then conclude C. Basis for computing a set of plausible extensions to the KB. Abduction : for a given rule p→q and evidence of q, infer p ∀x A(x) → B(x) A(c) then conclude B(c) eg. measles(x) → spot(x) if we notice spot, then infer (conclude) measles. 9/8/2021 Artificial Intelligence 5

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. AI (Artificial Intelligence) Inheritance Inheriting attribute values from a prototype descriptions. CWA : only objects that satisfy any predicate P are those that must. Problems : CWA is not true ≠ in the world. 9/8/2021 Artificial Intelligence 6

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. AI (Artificial Intelligence) MR Adding new axioms increases the amount of knowledge contained in the knowledge base. ∴ the set of facts & inferences can only grow larger I. e. they increase monotonically. NMR New facts might become known which contradicted & invalidated old knowledge. The old knowledge was retracted causing other dependent knowledge to become invalid, thereby req. further retractions. The retractions led to a shrinkage or nonmonotoric growth in the KB at times. 9/8/2021 Artificial Intelligence 7

Dept. Computer Science, Korea Univ. Intelligent Information System Lab. AI (Artificial Intelligence) TMS(Truth Maintenance System) To main consistency of the knowledge being used by the problem solver & not to perform any information function. P. 275 7. 2. 3 TMS is used to preserve the logical integrity of the conclusions of inference system How? To recompute support(justification) for items in KB whenever necessary. i. e. items of the KB are revised ( Keeping(saving) justifications for ∀ inference & recompute support for its conclusions with the new belief) 9/8/2021 Artificial Intelligence 8