0% found this document useful (0 votes)
668 views5 pages

Artificial Neural Networks - Short Answers

An artificial neural network (ANN) is a computing system designed to mimic the human brain. There are different types of learning in ANNs, including correlation learning where connections between neurons are strengthened or weakened based on whether they fire at the same or different times, competitive learning where neurons compete to be winners for an input pattern, supervised learning using labeled examples provided by a teacher, and unsupervised learning where natural groupings of unlabeled inputs are formed. ANNs have advantages like nonlinearity, input-output mapping, adaptivity, fault tolerance, and neurobiological analogy. They have applications including classification, clustering, pattern association, function approximation, and forecasting.

Uploaded by

Mmk Reddy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
668 views5 pages

Artificial Neural Networks - Short Answers

An artificial neural network (ANN) is a computing system designed to mimic the human brain. There are different types of learning in ANNs, including correlation learning where connections between neurons are strengthened or weakened based on whether they fire at the same or different times, competitive learning where neurons compete to be winners for an input pattern, supervised learning using labeled examples provided by a teacher, and unsupervised learning where natural groupings of unlabeled inputs are formed. ANNs have advantages like nonlinearity, input-output mapping, adaptivity, fault tolerance, and neurobiological analogy. They have applications including classification, clustering, pattern association, function approximation, and forecasting.

Uploaded by

Mmk Reddy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 5

Artificial Neural Networks

Short answer Questions and answers


1. Define artificial neural network.
An artificial neural network (ANN) is a computing system designed to simulate the way
the human brain analyzes and processes information.

2. Define correlation learning.


This is called Hebbian learning. Hebb’s rule (or Hebbian rule) states that “ When an
axon of cell A is near enough to excite a cell B and repeatedly or persistently takes place
in firing it, some growth process or metabolic change takes place in one or both cells
such that A's efficiency, as one of the cells firing B, is increased”
The connections between two neurons might be strengthened if the neurons fire at the
same time and might weaken if they fire at different times.
The strength of connections between neurons represent the correlation between their
outputs ∆ wi , j=c wi , j x i x j
where c is some small constant, denotes the strength of the connection from the jth node
to the ith node, and xi and xj are the activation levels of these nodes.

3. Define competitive learning.


When an input pattern is presented to a network, different nodes compete to be winners.
The competitive process involves self-excitation and mutual inhibition among nodes,
until a single winner emerges.

4. What is supervised learning?


Learning by examples (training set) provided by the teacher.

In supervised learning, a teacher is available to indicate whether

i) a system is performing correctly or


ii) to indicate a desired response or
iii) to validate the acceptability of a system's responses or
iv) to indicate the amount of error in system performance.

Ex: Classification
5. What is unsupervised learning?
In unsupervised learning there is no explicit teacher, and the system forms clusters or
natural groupings of the unlabeled input patterns.
Ex: Clustering

6. Define fully connected network


In this architecture every node is connected to every other node.
These connections may be
a) Excitatory (Positive Weights)
b) Inhibitory (Negative Weights)
c) Irrelevant (Almost zero weight)
In a fully connected asymmetric network, the connection from one node to another
may carry a different weight than the connection from the second node to the first
In symmetric network the weight is same in both the directions.
7. What are the advantages of ANN?
Nonlinearity: In ANN Interconnection of non-linear neurons. The nonlinearity is
distributed throughout.
Input-Output mapping: Learning with teacher or without teacher.
Adaptivity: Can adapt the free parameters to changes in the environment.
Evidential Response: Not only that the ANN gives the response based on the inputs, but
gives decision with confidence.
Fault Tolerance: Another advantage of ANN is its ability to fault tolerance. If one
particular connection in human nervous system is not functioning properly, we can’t
notice this change. If many interconnections fail there will be degradation in the
performance. How much degradation that is tolerated is called graceful degradation.
VLSI Implementability: Integration of artificial neurons. The entire human brain cannot
be modelled. Particular task/application can be implemented.
Neurobiological Analogy: The ability to model the human system

8. List the applications of ANN


1. Classification
2. Clustering
3. Vector Quantization
4. Pattern Association
5. Function Approximation
6. Forecasting

9. Define vector quantization.


For compressing huge input data into small number of weight vectors vector quantization
is used. Vector quantization is the process of dividing up space into several connected
regions called Voronoi regions. Each region is represented using a single vector called a
codebook vector.

10. What is a perceptron?


A perceptron to be a machine that learns using examples, to assign input vectors or
samples to different classes using a linear function of the inputs.

11. How non-numeric inputs are handled in perceptron training algorithm?


Example
1. The input dimension may be "color," and its values may range over the set {red, blue,
green, yellow}. By using binary vector with one component corresponding to each
color. The disadvantage with this method is as the dimensions of input increases,
vector size increase.

2. The weight function on the connection from the color node to the jth node at the next
level may be of the form

12. What is Backpropagation?


Error at a higher or outer layer of a multilayer network can be propagated backwards to
nodes at lower or inner layers of the network and the network weights are modified to
minimize the mean squared error between the desired and actual outputs of the

network

13. What kind of learning is used by backpropagation algorithm?


Supervised learning

14. What is adaline?


An adaptive linear element or Adaline, is a simple perceptron-like system that
accomplishes classification by modifying weights in such a way as to reduce the Mean
Square Error (MSE) at every iteration.

15. What is over training?


Excessive training on the training data sometimes decreases performance on the test data.
This is called overtraining.
16. How learning rate is selected in Backpropagation?
Training is generally commenced with randomly chosen initial weight values.
Typically, the weights chosen are small between -1.0 and 1.0 or -0.5 to +0.5, since larger
weight magnitudes may drive the output of layer 1 nodes to saturation, requiring large
amounts of training time
If the magnitudes of some of the inputs are much larger than others, random initialization
may bias the network to give much greater importance to inputs whose magnitudes
happen to be large. In such a situation, weights can be initialized as follows, so that the
net input to each hidden node is roughly of the same magnitude.

17. What is XOR problem?

Logical XOR function truth table is as shown in figure. Here 0s represent one
class and 1s another class. These samples are not linearly seperable. Cannot
be separated with a single line implemented with perceptron. This is called
XOR problem.

18. Give two applications of feedforward networks trained using Backpropagation algorithm.
1. Weaning from mechanically assisted ventilation: It is a two-class medical
classification problem for weaning patients from respiratory assistance.
2. Classification of myoelectric signals : It is a classification problem in which
myoelectric signals are distinguished.
3. Forecasting commodity prices: This is a financial forecasting problem, predicting
values for flour prices in three cities. (any commodity)
4. Controlling a gantry crane : This is a problem in automatic control, in which the
behavior of a gantry crane is to be understood and manipulated

5. What is Boosting?
Boosting is an approach which improves performance by systematic repeated application
of base learning algorithm such as back-propagation and the combining the results of
these iterations.

6. What is Bagging?

: In bagging a random sample is taken from the training set, and the base algorithm is
applied on the choosen sample to obtain a hypothesis. The process is repeated T times
and the final bagging hypothesis is calculated is calculated as

ht
H bagging   t 1
T

You might also like