Preliminary concepts Flashcards
What is reasoning (in the AI context)?
The formal manipulation
of symbols to reach a goal
What are the two types of knowledge reasoning within AI? Explain these two terms.
- Symbolic reasoning: It works by manipulating a collection of symbols to produce new knowledge. Transparant but limited.
- Sub-symbolic reasoning: It works by operating with numbers to produce new
knowledge. Often powerful but difficult to understand.
What two types of ‘sentences’ are used in Knowledge representation? Explain them.
- Atomic sentences represent simple facts or constants (John is a child of Sue);
- Conditional sentences represent complex facts (If X is a child of Y then Y is a parent of X)
What is logical entailment?
A type of thinking that determines whether an atomic sentence is true based on the given knowledge. If all statements in the antecedent are true, the query is true as well.
What is back-chaining?
A logical procedure that is used to establish if an atomic sentence is logically entailed. It can also be seen as working backward from the goal. Back-chaining works backwards from the consequent to the antecedent.
Example:
1. Look for “George is the father of Sue.” Nothing found
2. Look for “If…., then George is the father of Sue.” Found “If X is a child of Y and Y is male then Y is a father of X”. This matches for Y=george and X=sue.
3. Look for “Sue is a child of George” and look for “George is male”. Both found true
Return success
What is backtracking?
If a value leads to failure, back-chaining must go back to where the
answer was found and reconsider another answer. This procedure is
referred to as backtracking.
What are antecedents and consequents?
- Antecedents are the if… part in a conditional sentence
- Consequents are the then… part in a conditional sentence
What is the generate-and-test strategy?
We generate a value, if it is false, backtracking is used to generate and test a new value.
