2. Introduction:
Artificial Neural Networks are computational models
inspired by human brain, used to solve complex problems. This
paper is written to introduce artificial neural networks with new
comers from computers science researchers and developers. This
paper covers only those concepts from Biological Neural
Network which are compulsory for computer science field.BNN
have many other parts which are not covered here because of
unnecessity.To understand ANN, basics of BNN(nervous system)
should be clear.
3. Artificial Neural Network.
The idea of ANNs is based on the
belief that working of human brain by
making the right connections, can be
imitated using silicon and wires as living
neurons and dendrites.
The human brain is composed of 86
billion nerve cells called neurons. They
are connected to other thousand cells by
Axons.
4. Stimuli from external environment or inputs from sensory
organs are accepted by dendrites. These inputs create electric
impulses, which quickly travel through the neural network.
A neuron can then send the message to other neuron to handle
the issue or does not send it forward.
ANNs are composed of multiple nodes, which imitate
biological neurons of human brain. The neurons are connected
by links and they interact with each other. The nodes can take
input data and perform simple operations on the data.
5. The result of these operations is passed
to other neurons. The output at each node
is called its activation or node value.
Each link is associated with weight.
ANNs are capable of learning, which
takes place by altering weight values.
6. ANNs Working:
In the artificial neural network each arrow
represents a connection between two
neurons and indicates the pathway for the
flow of information.
Each connection has a weight, an integer
number that controls the signal between
the two neurons.
If the network generates a “well or not
well” output, there is no need to adjust the
weights. If the network generates a “poor
or undesired” output or an error, then the
system alters the weights in order to
improve subsequent results.
7. Artificial Neural Network Architecture:
An Artificial Neural Network is
defied as a data processing system
consisting of a large number of simple
highly interconnected processing elements
in an inspired by the structure of the
cerebral cortex of the brain.
8. An Artificial Neural Network
structure can be represented using a
directed graph. A graph G is an ordered
of 2 -tuple (V,E ) consisting of set of V
vertices and set of E edges. It is called
as digraph or directed graph.
Vertices is represented as neurons
(inputs, outputs) and the edges is
synaptic links.
9. Single Layer Feedforward Network:
The single layer feedforward
network contains two layers input
layer and output layer.
The input layer neurons receives the
input signals. The output layer
receives the output signals.
10. The synaptic links will carry their weight and connects to each
input and output neuron.
Such a network is a said to be feedforward in a type acyclic in
nature.
The input layer transmits the signals to the output layer. Hence
this is called as single layer feedforward network.
11. Multilayer feedforward network:
The multilayer feedforward network has
three layer input layer, output layer and
intermediate layer is called hidden layer.
Hidden layer is called hidden neurons
or hidden units.
12. One input layer, one output layer and two hidden layer .
The input layer neurons are linked to the hidden layer neurons
and weight is carried in their links is called input hidden layer
weight.
The hidden layer neurons are linked to the output layer neurons
and the weights are called as hidden output layer weights.
13. Recurrent networks:
These network differ from
feedforward network architecture in
the sense that there is atleast
feedback loop.
There could also be neurons with
self-feedback links the output of a
neuron is feedback into itself as
input.
14. Characteristics of neural network:
The neural network has mapping capabilities, that is they can
map input patterns to their associated output patterns.
The neural networks process the capability to generalize.
The Neural network are robust systems and are fault tolerant.
Recall all patterns from incomplete, or noisy patterns.
The neural network can process information in parallel, at
high speed, and in a distributed manner.
15. APPLICATIONS:
Airline Security Control.
Investment Management and Risk
Control.
Prediction of Thrift Failures.
Prediction of Stock Price Index.
OCR Systems.
Industrial Process Control.
Data Validation.
Risk Management.
Target Marketing.
Sales Forecasting.
Customer Research.
18. Supervised Learning.
In this learning the input pattern will
be trained the output pattern for the
target of the desired pattern.
We assume that a teacher will be
present in the class during the learning
process.
A comparison will be done between
the computed output and corrected
output to find the error.
The error can be change by network
parameter by the improvement of the
result performance.
19. Unsupervised learning:
In this learning method, the
target output is not presented in the
network that is a teacher will not be
presented in the class in the desired
pattern. So the system will be
discovered by its own knowledge
by its input patterns.
20. Reinforced learning:
In this method a teacher will be
present in the class by they won’t
correct the output only they will
indicate that the output is correct or
wrong.
In this learning method a reward will
be given for the correct answer and the
penalty will be given for the wrong
answer.
21. Hebbian learning:
Hebbian learning is based on correlative
weight adjustment.
The input-output pattern pairs(X i ,Y i)
are associated by the matrix W, known as
correlation matrix.
is transpose of output vector Yi.
22. Gradient descent learning:
This is based on the minimization of
error E defined in terms of weights
and the activation function of the
network.
If Wij is the weight update of the
link connecting ith and jth neuron of
the two layers Wij is defined as
23. Competitive learning:
The neurons which respond strongly to input stimuli have their
weights updated.
When an input pattern is presented, all neurons in the layer
compete and the winning neuron undergoes weight adjustment.
That “winner takes all “ strategy.
Stochastic learning:
In this the weights are adjusted in a probabilistic fashion.
24. Conclusion:
In this paper we discuss about the Artificial neural network,
methods and application of Artificial neural network. In this Artificial
neural network technology is developing day by day. It is useful for
all human that is it will reduce the time of learning and by this it
solves our learning problem. We can save more and more time and
money in any work. A process of learning speed will be increased
and so it is benefit for us. A improvement should be need for day by
day in our technology period. We can develop much more algorithms
and problems.