print version
Download
Report
Transcript print version
Dealing with Uncertainty
The need to deal with uncertainty arose in “expert systems”
Code expertise into a computer system
Example:
Medical diagnosis: MYCIN
Equipment failure diagnosis in a factory
Sample from MYCIN:
IF
The infection is primary-bacteremia AND
The site of the culture is one of the sterile sites AND
The suspected portal of entry is the gastrointestinal tract
THEN
There is suggestive evidence (70%) that the infection is bacteroid
Expert systems often have long chains
IF X THEN Y … IF Y THEN Z … IF Z THEN W …
If uncertainty is not handled correctly, errors build up, wrong diagnosis
Also, there may be dependencies, e.g. X and Y depend on each other
Leads to more errors…
Need a proper way to deal with uncertainty
How do Humans Deal with Uncertainty?
Not very well…
Consider a form of cancer which afflicts 0.8% of people (rare)
A lab has a test to detect the cancer
The test has a 98% chance to give an accurate result
Mr. Bloggs goes for the test
The result comes back positive
i.e. the test says he has cancer
What is the chance that he has the cancer?
28%
Afflicts experts too
Studies have shown: human experts thinking of likelihoods do not reason
like mathematical probability
A
Metastatic cancer
No Link
Increased total
serum count
B
C
Coma
D
Brain Tumour
E
Severe
headaches
A
Metastatic cancer
… …
… …
… …
Increased total
serum count
B
… …
C
Brain Tumour … …
… …
… …
… …
… …
Coma
Serum
count
Brain
Coma
tumour
Yes
Yes
Yes
D
E
Severe
headaches
95%
Brain
tumour
headache
No
94%
Yes
70%
No
Yes
29%
No
1%
No
No
0.1%
Inference in Belief Networks
Questions for a belief network:
Diagnosis
Work backwards from some evidence to a hypothesis
Causality
Work forwards from some hypothesis to likely evidence
Test a hypothesis, find likely symptoms
In general – mixed mode
Give values for some evidence variables
Ask about values of others
No other approach handles all these modes
Reasoning can take some time
Need to be careful to design network
Local structure: few connections
How Good are Belief Networks?
Relieves you from coding all possible dependencies
How many possibilities if full network?
Tools are available
Build network graphically
System handles mathematical probabilities
Case study:
Pathfinder a medical expert system
Assists pathologists with diagnosis of lymph-node diseases
Pathfinder is a pun
User enters initial findings
Pathfinder lists possible diseases
User can
Enter more findings
Ask pathfinder which findings would narrow possibilities
Pathfinder refines the diagnosis
Pathfinder version based on Belief Networks performs significantly better
than human pathologists