Download (p « q) 1 1 0

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Personal knowledge base wikipedia , lookup

History of artificial intelligence wikipedia , lookup

Embodied cognitive science wikipedia , lookup

Embodied language processing wikipedia , lookup

Knowledge representation and reasoning wikipedia , lookup

Transcript
“Victor Babes”
UNIVERSITY OF MEDICINE
AND PHARMACY
TIMISOARA
DEPARTMENT OF
MEDICAL INFORMATICS AND BIOPHYSICS
Medical Informatics Division
www.medinfo.umft.ro/dim
2004 / 2005
COURSE 12
MEDICAL DECISION
SUPPORT (II)
4. ELEMENTS OF LOGICS
4.1. GENERAL NOTIONS
• a) SENTENCE
– expressing ideas about different objects or
events features or causes
– TYPES: cognitive, interogative, imperative...
– TRUTH VALUE :
• True (T), False (F), Uncertain (?)
• b) LOGIC FORMS :
– notion, sentence, inference
– logic form of a declarative affirmative
sentence: S is P (S=subject, P=predicate)
• c) PRINCIPLES OF LOGICS
– identity principle
– non-contradiction principle
– excluded tertiary principle
– sufficient rationale principle
• d) CATEGORICAL SENTENCES:
– Universal - affirmative/negative
• All S are P. None S is P.
– Particular - affirmative, negative
• Some S are P. Some S are not P.
4.2. COMPOSED SENTENCES
• Applying an operator on one or two simple sentences
(unary or binary operators)
• Truth value of composed sentences:
• Negation
p
1
0
(not)
~p
0
1
• Conjunction
(p  q)
(and)
q
p
1
0
1
1
0
0
0
0
• Disjunction
(p  q)
(or)
q p
1
0
1
1
1
0
1
0
• Implication
(p  q)
(if…then...)
q p
1
0
1
1
0
0
1
1
• Exclusive
q p
disjunction
1
(or…or...) XOR0
1
1
0
0
0
1
• Equivalence
(p  q)
(if and only if)
1
1
0
0
0
1
q p
1
0
4.3. LOGICAL INFERENCE
• Structure: premises (2 sen)
conclusion
• Modus
ponens
(p  q)
p
q
• Modus
tollens
(p  q)
~q
~p
• Sillogism
(p  q)
(q  r)
(p  r)
4.4. Ex: PROLOG LANGUAGE
domains
person, activity = symbol
predicates
likes (person, activity)
clauses
likes (ellen, tennis)
likes (tom, baseball)
likes (bill, X) if likes (tom, X)
RUN: goal: likes (bill, baseball)
TRUE
5. HEURISTIC METHODS: EXPERT SYSTEMS
5.1. SCHEME
5.2. COMPONENTS
• a) COGNITIVE SYSTEM
–KNOWLEDGE BASE: 3 levels
• FACTUAL
• CONCEPTUAL
• (meta-knowledge)
• b) REASONING SYSTEM
(inference machine)
• c) COMMUNICATION SYSTEM
(user interface)
• d) EXPLANATORY SYSTEM
• [ e) Meta-resolutive system
– to check if results are valid and
reasoning is adequate) ]
5.3. COGNITIVE SYSTEM AND
KNOWLEDGE BASE
• Medical knowledge extraction:
– general knowledge
– knowledge from clinical experience
• Medical knowledge formalization
• KB may be extended
• Self-completion (PROLOG language –
appropriate for “Artificial Intelligence)
• May be used either for decision support or for
educational purposes)
• 5.4. REASONING SYSTEM
– leading position - drives the dialogue
– may accept statistical procedures
• 5.5. EXPLANATORY SYSTEM
• shows the reasoning ‘trace’
– educational purposes
• 5.6. COMMUNICATION SYSTEM
– natural language use
5.7. MEDICAL Expert Systems
•
•
•
•
•
•
•
•
MYCIN - bacterial infections
PUFF - pulmonary deseases
HEADMED - neuro-psychiatry
CASNET - ophtalmology
VM -’ventilator monitor’
INTERNIST - internal medicine
TROPICAID - tropical diseases
Domain-independent E.S.
6. CLASSIFICATION QUALITY
ASSESSMENT
a) Classification table
Classifier
T+ TD+ N11 N12 R1
Real D- N
21 N22 R2
C 1 C2 N
b) PARAMETERS
•
•
•
•
true positive = TP = N11
false negative = FN = N12
false positive = FP = N21
true negative = TN = N22
• sensitivity = p(T+/D+) = N11 / R1
– (capacity to detect the diseased)
• specificity = p(T-/D-) = N22 / R2
– (capacity to reject the non-diseased)
• positive predicitve value
ppv = p(D+/T+) = N11 / C1
• negative predictive value
npv = p(D-/T-) = N22 / C2
Global indicators
• accuracy = (N11 + N22) / N
• classification error rate = (N12 + N21) / N
Threshold value
d) ROC Curve SN = f(1 – SP)
Receiver – Operator Characteristic
SN si SP vary inversly proportional
c) Example
• A study was carried on a population of 3000
individuals. 500 of them had virosis during
last year. Our computer program was able to
detect 480 of them, but gave the same
diagnosis to another 50 persons. Compute:
• number of false positives and false negatives
• sensitivity and specificity
• accuracy and classification error rate
• positive and negative predictive rate
End