Computational Intelligence Lecture 1 JFBaldwin

habitualparathyroidsΤεχνίτη Νοημοσύνη και Ρομποτική

7 Νοε 2013 (πριν από 3 χρόνια και 5 μήνες)

105 εμφανίσεις

Computational
Intelligence
Lecture 1
J.F.Baldwin
Induction
Set of +ve and -ve
Examples for a given concept
Knowledge
Rules for concept
or
other form of
knowledge
representation
specific
general
Ace of spades - good
Queen of diamonds - good
King of hearts - good
Jack of clubs - good
10 of diamonds - bad
6 of clubs - bad
3 of hearts - bad
5 of spades - bad
Example Set
Induced Rule:
card good IF card is Royal else bad
Background Knowledge: Ace, King, Queen, Jack are
Royal Cards
Background Knowledge
using background knowledge
No guarantee of validity
Examples
1 2 3 4 ?What is next number in sequence?
etc
Pictures of
male faces
etc
Pictures of
female faces
male or
female face
What features should we use for this classification
What weights of importance should we give to these features
General Points
Features should discriminate between +ve and -ve concepts
Many rules will satisfy examples
Science attempts to find the simplest rules
We can generalise to classification problems with more than two
classes
We should be able to make predictions as well as classifications
Why do we need Induction
Data is Everywhere - Databases
Science
Medicine
Engineering
Finance
Marketing
Social Welfare
Education
Trade
We want knowledge NOT data Bases
What is Knowledge ?
Should we use propositional logic ?
Should we use predicate logic ?
What confidence can we have in our generalisations ?
What tests can we apply ?
How can we discover the knowledge automatically from
the large volume data ?
How can we update our knowledge as we discover mistakes ?
Can we have a logic of induction as we have a logic of
deduction ?
An Example to think about
Good L
Bad L
unknown
twice
twice
Example Set
How can you classify these
Remember : There is no correct solution
INTELLIGENT FUZZY DATA BROWSER
INTELLIGENT FUZZY DATA BROWSER
INTELLIGENT FUZZY DATA BROWSER
INTELLIGENT FUZZY DATA BROWSER
FRIL
FRIL
FRIL
FRIL
Relevant Data Collection
Net
Databases
Intelligent Agents
Generalisation & Summarization
Inference
Query & Solution Interpretation
Human / Computer Interface
Knowledge
Base
Integrity
Checker
Conversation
Supervisor
Fuzzy
Inductive
Logic
Numeric, Text, Audio, Graphics
Feature Extraction
DATA
INFORMATION
KNOWLEDGE
THEORIES
Soft Computing
Soft Computing
Soft Computing
Fuzzy Set Theory
&
Fuzzy Logic
&
Support logic
Probability Theory
Neural Net
Theory
Genetic
Programming
Evolutionary
Programming
Probability Fuzzy
Mass Assignment Theory
?
Others
Inductive
Logic
Another example to think
about
card 1 - bad
card 2 - good
card 3 - good
card 4 - bad
card 5 - bad
card 6 - good
card 8 - ???
card 7 - ???
predict card
type
INDUCE
Some Applications
Learning rules from example sets
to classify under-water sounds
to act as an electronic nose
to identify speech
to label a picture
to control a process
to predict the stock market share prices
to classify credit worthiness
to diagnose faults and medical disorders
to approximate a function
to predict the weather
to design an expert system
See notes for geology example
Data May not be Perfect
Data may be noisy
There may be missing data
Data may be vague in places
The Inductivist Controversy
H --> R
H
R
Logic Deduction
H --> R
R
H  Â H
Induction is not valid
We cannot know anything for sure
H --> R
 R
 H
Falsification
We can falsify an Hypothesis
H --> R
H
H --> R
R
Induction
Deduction
Induction is, in a sense, the inverse of deduction
Machine Learning
Bacon - induction for scientific discovery can be mechanical

Popper - cannot induce, only falsify
Scientific method consists of generating hypotheses
by some creative process and then attempting to falsify
When a hypothesisest.
Example: relativity from NewtonÕs theory
Turing - logic and practice are important for machine learning
Valiant - probably approximately correct learning, PAC
TODAY
Machine Learning from examples set +background
knowledge has been found to be successfulrÕs Falsification
Ideas
We can use background knowledge and mechanical
means to generate hypotheses
The process still relies to some extent on choosing
reasonable features
Some Bacon, some Popper, some Turing, Some Valiant