Knowledge and Logical Agents (needs refinement)

Systems that think rationally, it will laconically infer

What is an Expert System

• AI gets knowledge from an experienced human - ML gets data from data analysis not human

Task types

Diagnosis tasks - fixing a device, medical diagnosis

Assessment tasks - offering loads, deciding insurance rates

Develop an expert system

• Know to get knowledge from a human

• Analyse usage:
  • Forward chaining - start with facts (effect), to derive conclusions (causes)
  • Backward chaining - start with conclusion that needs to be proven, end with facts to prove it

• Design a well know methodology aka use CommonKADS

Knowledgeable, logic, inference

Knowledgeable (KB) - set of facts that can be used to answer a query

Logic - KB syntax

Inference - Is a formal method to conclude or prove facts from existing facts

Other ways of repesenting KB syntax beyond logic

Ontology - represents the relationships between classes or properties of things

Semantic Network - repesents the specific objects (I THINK)

Types of Logics

Types:

• Bi-value

• Multi-valued logic

• Probabilistic logic

• Fizzy Logic

Propositional Logic 0-order

Uses: Bi-value/boolean logic

Operations:

• NOT

• AND

• OR

• IF/THEN

Predicate Logic First-Order Logic (FOL)

Operations:

• FOR EACH day

• RELATIONSHIPS / FUNCTIONS: partOf(), has(), brotherOf()

• VARIABLES: X, Y, Z

• QUANTIFIERS: All, there is (ADD SYMBOLS)
  • Example:
    There is X that likes ice cream
  • Far all X, if X is at NTU then X is cool

Higher-order logics (not really covered)

Comparisons or equivalents between boolean logic

MUST KNOW THE FOLLOWING:

A AND B == NOT A OR B

Inference

• Deduction theorem: P infers Q if and only if P AND Q is TRUE

• But if P AND Q is FALSE we cant say that Q is FALSE, aka it cant disprove a fact, it can only prove facts

• You can disprove Q, by proving NOT Q

• However, KB might not be enough, to prove eather