Enhancing the Explanatory Power of Intelligent, Model

Download Report

Transcript Enhancing the Explanatory Power of Intelligent, Model

Artificial Intelligence
Robbie Nakatsu
AIMS 2710
Artificial Intelligence--AI
The techniques and software that enable
.
computers to mimic human behavior
in various ways. A major thrust in this
field is to develop computer functions
associated with human intelligence.
Some Types of AI
• Expert systems
• Natural language processing
• Machine learning
• Robotics
• Intelligent Agents
• Logic reasoning
AI is like magic!
“Any sufficiently advanced technology is
indistinguishable from magic.”
--Arthur C. Clark, 1962
An Expert System
is an AI program that emulates the decisionmaking ability of a human expert
An expert system captures expertise from a
human expert and applies it to a problem.
An Expert System can perform
diagnostic and prescriptive tasks like:




Auditing and tax planning
Diagnosing illnesses
Commercial loan decisions
Determining the cause of machine failure
What is the difference between a
diagnostic and prescriptive task?
People In An Expert System



Domain Expert - the person who knows how to solve
the problem without the aid of IT.
Knowledge Engineer - the person who works with
domain experts to capture knowledge they possess.
The knowledge engineer builds the expert system.
End User - the person who uses the expert system
to solve a problem.
Components of an Expert
System
Knowledge
Base
facts
User Interface
User
recommendations
Inference
Engine
Working
Memory
Components Defined
•
KNOWLEDGE BASE - stores the domain expertise (e.g.,
a collection of If-Then rules).
•
INFERENCE ENGINE - processes the domain expertise
and your problem facts to reach a conclusion.
•
•
WORKING MEMORY – short term memory of the expert
system; contains all the facts (initial facts as well as new
facts).
USER INTERFACE – part of the expert system that you
use to run a consultation.
Representing Expertise as a Collection
of Rules
IF the light is green THEN
Go through the intersection
If the light is red THEN
STOP
If the light is yellow AND there is time to go through intersection
before the light turns red THEN
Go through the intersection
If the light is yellow AND there is not time to go through
intersection before the light turns red THEN
STOP
A more complex example
IF 1. The infection that requires therapy is meningitis AND
2. The patient has evidence of serious skin or soft
tissue infection AND
3. Organisms were not seen on the stain of the
culture AND
4. The type of infection is bacterial
THEN
There is evidence that the organism that might be causing the
infection is Staphylococcus coagpos (0.75) or Streptococcus
(0.5)
Inference Engine
It is the part of the Expert System that processes the
problem facts and searches for rules in the
knowledge base to reach a final recommendation for
a user. Two inferencing strategies :
Forward Chaining is a data-driven approach in which
you start with the initial problem facts, and then try
to draw conclusions from them using the rules of the
knowledge base.
Backward Chaining is a goal-driven approach in
which you start with some kind of expectation of
what is to occur, or hypothesis, and then find rules
that either support or contradict your hypothesis.
Illustrating Forward and
Backward Chaining
Knowledge Base
R1: IF A and C, THEN E
R2: IF D and C, THEN F
R3: IF B and E, THEN F
R4: IF B, THEN C
R5: IF F, THEN G
Two Problems:
1.
Forward Chain: Assume B and D
2.
Backward Chain: Prove or
disprove G, and assume A and B
Expert System Opportunities
Any activity where human experts are
overburdened, undersupplied, or expensive are
good candidates for ES.
•
•
Expertise might be scarce in some organizations
(can propagate the expertise through the use of
an ES).
An ES might also be used to enhance the role of
an expert by providing the necessary assistance.
Benefits of Expert Systems
•
•
•
•
•
•
Increased output and productivity
Reduced costs, including decreased personnel
required
Fewer errors
Better and more consistent decision-making
Knowledge transfer to remote locations
Formalization of organizational knowledge
Questions for thought


What are some problems and
limitations of expert systems?
Can expert systems solve all kinds of
problems?
Machine Learning
Field of study that gives computers the
ability to learn without being explicitly
programmed (Arthur Samuel, 1959)
Example: Checkers playing
program that sees tens of
thousands of examples of
board positions, and learn
over time what the good
positions are.
Learning from Data (see video)

Regression problems (predicting continuous-valued
outcome)
 Predicting price of home from its size
 Predicting price of home based on multiple variables
(size, year built, location, condition of the building, etc)

Classification problems (predicting discrete-valued
outcome)
 Determining malignancy of a tumor based on its size
 Determining malignancy of a tumor based on multiple
variables (size, age of patient, uniformity of tumor, etc.)
Some Examples of Machine Learning




A credit card company wants to predict whether a credit
card transaction is fraudulent or not.
A company that sells ice cream wants to predict how much
ice cream to produce over the summer months (June –
August).
A software company wants to design an email spam filter
to predict whether an email is spam or not.
A marketing researcher has customer data and wants to
predict who among the customers are the most profitable.
Which of the above are classification problems and which are
regression problems?
Housing price prediction.
400
300
Price ($)
200
in 1000’s
100
0
0
500
1000
1500
2000
2500
Size in feet2
Supervised Learning
“right answers” given
Regression: Predict
continuous valued output
(price)
Breast cancer (malignant, benign)
Classification
Discrete
valued output
(0 or 1)
1(Y)
Malignant?
0(N)
Tumor Size
Tumor Size
- Clump Thickness
- Uniformity of Cell Size
- Uniformity of Cell
Shape
…
Age
Tumor Size
A Neural Network
is an artificial intelligence system which is
capable of learning to recognize patterns
and relationships in the data it processes.
A neural network simulates the human ability to
classify things based on the experience of seeing
many examples.
A Neural Network can perform
pattern recognition tasks like:




Detecting anomalies in human tissue that may
signify disease
Reading handwriting
Speech recognition
Detecting abnormal patterns in
electrocardiographs
An Intelligent Agent
is an artificial intelligence system that can move
around your computer or network performing
repetitive tasks independently, adapting itself to
your preferences.
An intelligent agent is like a travel agent in that it
performs tasks that you stipulate.
Examples

Intelligent search engines


Search engines that know who you are,
your preferences, where you are, who your
friends are, etc.
Personal assistants


Check and filter your e-mails
Search the web and collect important news
items for you
Intelligent Agent Characteristics



Autonomy
Adaptivity
Sociability
Recap and Summary







Types of decisions
Decision Support Systems
OLAP (online analytical processing)
Supporting groups with technology
Expert Systems
Machine Learning
Intelligent Agents