Transcript Aspects of Artificial Intelligence

Aspects of Artificial Intelligence
Himadri Barman
Defining AI
Artificial Intelligence
(AI) is a branch of
computer science concerned with the study and
creation of computer systems that exhibit some
form of intelligence. It means
• systems that learn new concepts and tasks
• systems that can reason and draw useful
conclusions about the world around us
• systems that can understand a natural language or
perceive and comprehend a visual scene
• systems that perform other types of feats that
require human types of intelligence
2
Business and AI
• To store information in an active form as
organizational memory, creating an organizational
knowledge base that many employees can examine
and preserving expertise that might be lost when an
acknowledged expert leaves the firm
• To create a mechanism that is not subject to human
feelings, such as fatigue and worry. This may be
especially useful when jobs may be environmentally,
physically or mentally dangerous to humans. These
systems may also be useful advisers in times of
crisis
3
• To eliminate routine and unsatisfying jobs held by
people
• To enhance the organization’s knowledge base by
generating solutions to specific problems that are
too massive and complex to be analyzed by human
beings in a short period of time
• Attempt to solve problems with no apparent solution
4
Fuzzy Sets
• Classical Sets - binary membership
(absolute membership - either yes or no)
• Fuzzy Sets - allows partial membership mapping into the interval [0,1]
5
Fuzzy Logic
• A form of knowledge representation suitable
for notions that cannot be defined precisely,
but which depend upon their contexts.
• Fuzzy logic generalizes the notion of truth
values in classical logic (i.e., true or false)
into a matter of degree. A statement in fuzzy
logic, thus may be partially true between 0
(false) and 1 (true). I've grayed to a
considerable extent, but not entirely grayed
may be represented by assigning a value
(say 0.6) the logic grayed (Himadri)
[Technically grayed will be called a predicate
and Himadri will be an argument]
6
Fuzzy Logic Example
• The problem
– Change the speed of a heater fan, based on the room
temperature and humidity.
• A temperature control system has four settings
– Cold, Cool, Warm, and Hot
• Humidity can be defined by:
– Low, Medium, and High
• Using this we can define the fuzzy set.
7
Biological Neurons
A neuron consists of a cell body, one axon, and many
dendrites. Dendrites receive inputs from axons of other
neurons via excitation or inhibition synapses. Synapses are
narrow gaps through which Axons influence dendrites
8
Artificial Neural Networks
• An ANN typically consists of simulated neurons.
• Each link is associated with a weight. The nature
and strength of one node’s influence on the
other is determined by the weight
• Each node combines the separate influences
received on its input links into an overall
influence using an activation function (most often
a threshold function is used corresponding to a
threshold value)
9
An Artificial Neuron
The artificial neuron (perceptron) receives inputs, processes inputs
and delivers a single output.
10
ANN Example
Adapted from Winston, P H, Artificial Intelligence, 3rd Edition, Pearson
Robert
Ronald
Romeo
H1
Hidden Node
Output Not observable
0.5
H3
Acquaintances
1.5
All other links weight is 1
-1.0
H4
John
James
0.5
-1.0
-1.5
Siblings
Jack
H2
Hidden Node
ANN recognizing siblings and acquaintances
11
ANN Applications
•
•
•
•
•
•
Tax form processing to identify tax fraud
Enhancing auditing by finding irregularities
Bankruptcy prediction
Customer credit scoring
Loan approvals
Credit card approval and fraud detection
12
• Financial prediction
• Energy forecasting
• Computer access security (intrusion
detection and classification of attacks)
• Fraud
detection
in
mobile
telecommunication networks
13
Benefits of ANNs
• Usefulness
for
pattern
recognition,
classification, generalization, abstraction and
interpretation of incomplete and noisy inputs.
(e.g.
handwriting
recognition,
image
recognition, voice and speech recognition,
weather forecasting).
14
• Providing some human characteristics to
problem solving that are difficult to
simulate using the logical, analytical
techniques of expert systems and
standard software technologies. (e.g.
financial applications).
15
• Ability to solve new kinds of problems.
ANNs are particularly effective at solving
problems whose solutions are difficult, if
not impossible, to define. This opened up
a new range of decision support
applications formerly either difficult or
impossible to computerize.
16
• Robustness. ANNs tend to be more robust
than their conventional counterparts. They
have the ability to cope with incomplete or
fuzzy data. ANNs can be very tolerant of
faults if properly implemented.
17
• Fast processing speed. Because they
consist of a large number of massively
interconnected processing units, all
operating in parallel on the same problem,
ANNs can potentially operate at
considerable speed (when implemented
on parallel processors).
18
• Flexibility and ease of maintenance. ANNs
are very flexible in adapting their behavior
to new and changing environments. They
are also easier to maintain, with some
having the ability to learn from experience
to improve their own performance.
19
Limitations of ANN
• ANNs do not produce an explicit model
even though new cases can be fed into it
and new results obtained.
• ANNs lack explanation capabilities.
Justifications for results is difficult to obtain
because the connection weights usually
do not have obvious interpretations.
20
Genetic Algorithms
GAs (also called Adaptive Computations) refer to
a variety of problem-solving techniques that are
conceptually based on the method that living
organisms use to adapt to their environments –
the process of evolution. They are programmed to
work the way populations solve problems – by
changing and reorganizing their component parts
using processes such as reproduction, mutation
and natural selection. GAs breed programs that
solve problems even when no person can fully
understand their structure.
21
1.
2.
3.
4.
5.
6.
7.
Initialize: Randomly generate a population of N
binary strings (chromosomes) where each string
represents a possible solution
Evaluation: Assign a fitness value to each string
which says how that solution is fit for the given
problem
Selection: Select n strings on the basis of fitness
Perform Crossover operation
Do Evaluation and Selection
Apply Mutation operation in the population of
strings
Repeat steps 2 to 6 till termination criteria satisfied
22
GA Example
Objective Function: Y=x2 – x ; x  (0,1, 2, …. 15)
Maximize the objective function
Total number of possible solutions or values of x = 16 = 24
So , a binary string of length l =4 is enough to represent a solution
Initialization
Let N = 4, randomly generate a population of 4 binary strings
0011
0101
1001
1000
23
GA Example
Evaluation
Assign a fitness value to each binary string
No.
Total
String (x)
X
Y=x2 – x
f = y / (∑yi)
1
0010
2
2
0.013
2
0101
5
20
0.133
3
1001
9
72
0.480
4
1000
8
56
0.374
150
1.000
Selection
Higher the fitness values, the more chance of being selected. Most likely
string 0010 won’t survive
24
GA Example
Crossover
Single point crossover
0101
1000
0100
1001
Double point crossover
11001011
10010001
11010011
10001001
Mutation
0101
0111
25