An Introduction to Artificial Intelligence and Knowledge Engineering

1
An Introduction to Artificial Intelligence and
KnowledgeEngineering
N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
2
Sub-topics

• Introduction to the AI paradigms (1.1 pp. 1-3)
• Heuristic problem solving (1.2 pp. 3-9)
• Genetic algorithms and evolutionary
  
  programming
  (1.2.3 pp. 9-14)
• Expert systems (1.3.1 pp. 14-15)
• Fuzzy systems (1.3.2 pp. 15-17)
• Neural networks (1.3.3 pp. 17-19)
• Hybrid systems (1.3.4 1.9, pp. 65-68)

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
3
Introduction to the AI Paradigms

• AI objectives
• to develop methods and systems for solving
  problems, usually solved through intellectual
  activity of humans, eg. image recognition
  language and speech processing planning,
  prediction, etc., thus enhancing the computer
  information systems
• to improve our understanding on how the human
  brain works

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
4
Introduction to the AI Paradigms (cont)

• AI directions
• developing methods and systems for solving AI
  problems without following the way the humans do
  (expert systems)
• developing methods and systems for solving AI
  problems through modelling the human way of
  thinking, or the way the brain works (neural
  networks)
• AI paradigms
• symbolic or sub-symbolic (connectionist)

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
5
Heuristic Problem Solving

• Figure 1.1
• Heuristics as means of obtaining restricted
  projections from the domain space D into the
  solution space S.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
6
Heuristic Problem Solving (cont)

• Figure 1.2
• (a) Ill-informed and (b) well-informed
  heuristics. They are represented as patches' in
  the problem space. The patches have different
  forms (usually quadrilateral) depending on the
  way of representing the heuristics in a computer
  program.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
7
Heuristic Problem Solving (cont)

• Figure 1.3
• The problem knowledge maps the domain space into
  the solution space and approximates the objective
  (goal) function (a) a general case (b) two
  dimensional case.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
8
Genetic Algorithms and Evolutionary Programming

• Gene
• Chromosome
• Population
• Crossover
• Mutation
• Fitness function
• Selection

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
9
Genetic Algorithms and Evolutionary Programming...

• Figure 1.4 An outline of a genetic algorithm

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
10
Genetic Algorithms and Evolutionary Programming...

• Figure 1.5 A graphical representation of a
  genetic algorithm

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
11
Genetic Algorithms and Evolutionary Programming...
Second player Answer (according to the criterion
001010) Produced variants (individuals) A)
010101 1 B) 111101 1 C)
011011 4 D) 101100 3
Using the criterion, the best ones are chosen
- C (mother) and D (father). Mating
New variants Evaluation C) 011011 E)
011100 3 D) 101100 F) 101011
4 C) 011011 G) 011000 4
D) 101100 H) 101111
3 Selection of F (mother) and G
(father) Mating New
variants Evaluation F) 101011 H)
111000 3 G) 011000 I) 001011
5 F) 101011 J)
101000 4 G) 011000 K)
011011 4 Selection of I (mother) and J
(father) Mating New variants
Evaluation I) 001011 L) 001000
5 J) 101000 M) 101011 4 I)
001011 N) 001010 6 (success)
END J) 101000 O) 101001 3

• Figure 1.6 An example of a genetic algorithm
  applied to the game "guess the number"

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
12
Expert Systems

• Expert systems are knowledge-based systems which
  contain expert knowledge and can provide an
  expertise, similar to the one provided by an
  expert in a restricted application area. For
  example, an expert system for diagnosis of cars
  has a knowledge base containing rules for
  checking a car and finding faulty elements, as it
  would be done by a specialised engineer.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
13
Expert Systems...

• Figure 1.7 The two sides of an expert system

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
14
Expert Systems...

• An expert system consists of the following main
  blocks
• knowledge base
• data base
• inference engine
• explanation module
• user interface
• knowledge acquisition module.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
15
Expert Systems...

• Example of an Expert System
• Rule 1
• IF (CScore is high) and (CRatio is good) and
  (CCredit is good)
• then (Decision is approve)
• Rule 2
• IF (CScore is low) and (Cratio is bad) or
  (CCredit is bad)
• then (Decision is disapprove)

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
16
Fuzzy Systems

• fuzzy sets
• fuzzy input and output variables
• fuzzy rules
• fuzzy inference mechanism

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
17
Fuzzy Systems...

• Figure 3.1 Membership functions representing
  three fuzzy sets for the variable "height".

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
18
Fuzzy Systems...

• Figure 3.2 Representing crisp and fuzzy sets as
  subsets of a domain (universe) U.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
19
Fuzzy Systems...

• Figure 1.8 A simple fuzzy rule for the smoker
  and the risk of cancer case example.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
20
Fuzzy Systems...

• Figure 3.16
• (a) Membership functions for fuzzy sets for the
  Smoker and the Risk of Cancer case example. (b)
  The Rc implication relation "heavy smoker gt
  high risk of cancer" in a matrix form.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
21
Neural Networks

• neural network structure
• learning
• generalization

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
22
Neural Networks...

• Figure 4.1
• A structure of a typical biological neuron. It
  has many inputs (in) and one output (out). The
  connections between neurons are realized in the
  synapses.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
23
Neural Networks...

• Figure 4.2
• A model of an artificial neuron.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
24
Neural Networks...

• Figure 4.5
• A simple neural network with 4 input nodes, two
  intermediate nodes and one output node. The
  connection weights are shown, presumably a
  result of training. The activation value of node
  n5 is shown too.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
25
Hybrid Systems

• Hybrid systems - A general approach to knowledge
  engineering
• Figure 1.37
• Different "pathways" can be used for knowledge
  engineering and problem solving to map the domain
  space into the solution space.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996
26
Hybrid Systems...

• Figure 1.38
• Usability of different methods for knowledge
  engineering and problem solving depending on
  availability of data and expertise (theories) on
  a problem.

N. Kasabov, Foundations of Neural Networks, Fuzzy
Systems, and Knowledge Engineering, MIT Press,
1996