Statistics Flashcards
(10 cards)
The fundamental notion is Bayesian statistics is that of…
Conditional probability
P(H|E)
The probability of a hypothesis H being true based on evidence E
When is Bayes rule used in the real world
Search and rescue
Fault diagnosis in engineering
Medical diagnosis
Email services
Fraud detection
What are knowledge based systems
A computer program that uses a knowledge base to solve problems
Hard coded (we put the information in and computer uses information)
Information is stored in such a way that the system can reason with it
Domain specific - narrow range of knowledge
Top down
What are two types of knowledge based systems
Expert systems
A list a rules
Cased based systems
A list of cases
What are the three steps of knowledge based systems
Acquire knowledge - transfer expertise from knowledge source to a program
Represent knowledge - symbolic encoding of propositions
Apply the knowledge, ie reasoning - deduce logical consequences: to produce new propositions
What is the bottleneck issue with knowledge based systems
Human knowledge is complex, unstructured and usually ill formatted
Experts may have subjective knowledge that may conflict
There may be missing knowledge
Knowledge acquisition/engineering three steps (Eliza example)
Eliza is a mock therapist
Define task
Build domain vocabulary
Develop a model of the reasoning and how it is applied (flow charts and decision tress often used)
How to represent natural language into a knowledge based
Predicate logic
A symbolic formal language to represent knowledge
Expressive to derive new knowledge from old, though mathematical deduction
What is the resolution rule (Knowledge based system reasoning) method, requirement and steps (4)
An inference method which infers new knowledge
Method: shows the counter-example leads to a conflict in the knowledge based (becomes empty)
Requirement: knowledge base is in Conjunctive Normal Form (CNF) ie propositions with clauses connected by v
Convert knowledge base to CNF
Add negated statement to knowledge base
Unification - replace variables by a concrete instance
Resolution - iteratively examining contradictions in the KB until an empty clause is resolved (the negated statement must be false) or no more contradictions exist
Bayes Rule
P(H|E) = (P(E|H)P(H))/P(E)
P(H|E) - the probability that hypothesis H is true given evidence E
P(E|H) - the probability that we will observe evidence E given that hypotheses H is true
P(H) - the a priori probability that hypothesis H is true
P(E) - the probability of observing E = P(E|H1)P(H1) + P(E|H2)P(H2) + … + P(E|Hn)*P(Hn)
