Uttar Pradesh Technical University (UPTU) 2007 B.Tech Information Technology Neutral Networks - Question Paper
Printed Pages : 4 IT - 031
(Following Paper ID and Roll No. to be filled in your Answer Book)
Roll No. | | | | | | | | | 1 | B. Tech.
PAPER ID : 1052
(SEM. VIII) EXAMINATION, 2006-07 NEURAL NETWORKS
Time : 3 Hours] [Total Marks : 100
Note : Attempt all questions. Each question carries 20 marks. Write suitable assumptions, if any.
1 Attempt any two parts :
(a) Discuss analogy between biological and 10 artificial neural network.
(b) What do you understand by the terms : 3+3+4
(1) Learning
(2) Generalization
(3) Function approximation in context of ANN.
(c) With help of a suitable diagram discuss 6+4 functioning of a simple artificial neuron. Explain how the functionality is affected if two such neuron are connected in series.
2 Attempt any two parts :
(a) Discuss the functioning of a perceptron. Consider 10 the classification problem defined below
{/?1=[-1, l], =l} {/,2=[0 0] *2=l}
~l] *3=l} {4=[ 0] *4=}
{5 =t *] *5 =} p. >i-th input t- i-th target
Design a single neuron perceptron to solve this problem. You need to describe architecture of perceptron, its decision boundary.
(b) Discuss back propagation algorithm for a 10 multilayer network.
(c) Discuss the functioning of recurrent networks. 6+4 What are strengths and limitations of such network.
3 Attempt any two parts :
(a) Consider LVQ network shown below
w
w
5x1
5x1
3x1
3x1
2x1
:w'-P
ni
[ 1 i = neuron with largest net input
a; = i 1 0 otherwise
2 2 , a =w a
0 0
|
1 |
0 |
0 |
0 |
0 |
|
0 |
1 |
0 |
0 |
0 |
|
0 |
0 |
1 |
1 |
1 |
Determine regions of each class and draw a graph to illustrate this.
(b) Write short notes on 4+6
(1) Vector quantization
(2) Unsupervised learning and
its relative merits and limitations with espect to supervised learning.
(c) Discuss Hebb rule in context of : 4+6
(1) Supervised learning
(2) Unsupervised learning.
4 Attempt any two parts :
(a) What do you understand by associative networks ? In this context explain working of a Hopfield network.
(b) Describe brain-in-a-box model. Compare it with a recurrent network.
(c) Derive expression for Boltzmann learning rule. 10
5 Attempt any two parts :
(a) Define gradient. Using steepest descent rule to 10 the following function
determine first three points of trajectory starting from
= [0.5, 0.5]
(b) Discuss the role of 2+4+4
(1) Selection
(2) Cross-over
(3) Mutation
in context of genetic algorithm.
(c) Write an algorithm to implement simulated 10 annealing.
V-1052] 4 [ 960 ]
Printed Pages : 4 IT - 031
(Following Paper ID and Roll No. to be filled in your Answer Book)
Roll No. | | | | | | | | | 1 | B. Tech.
PAPER ID : 1052
(SEM. VIII) EXAMINATION, 2006-07 NEURAL NETWORKS
Time : 3 Hours] [Total Marks : 100
Note : Attempt all questions. Each question carries 20 marks. Write suitable assumptions, if any.
1 Attempt any two parts :
(a) Discuss analogy between biological and 10 artificial neural network.
(b) What do you understand by the terms : 3+3+4
(1) Learning
(2) Generalization
(3) Function approximation in context of ANN.
(c) With help of a suitable diagram discuss 6+4 functioning of a simple artificial neuron. Explain how the functionality is affected if two such neuron are connected in series.
2 Attempt any two parts :
(a) Discuss the functioning of a perceptron. Consider 10 the classification problem defined below
{/?1=[-1, l], =l} {/,2=[0 0] *2=l}
~l] *3=l} {4=[ 0] *4=}
{5 =[ *] *5 =} p. >i-th input t- i-th target
Design a single neuron perceptron to solve this problem. You need to describe architecture of perceptron, its decision boundary.
(b) Discuss back propagation algorithm for a 10 multilayer network.
(c) Discuss the functioning of recurrent networks. 6+4 What are strengths and limitations of such network.
3 Attempt any two parts :
(a) Consider LVQ network shown below
w
w
5x1
5x1
3x1
3x1
2x1
:w'-P
ni
[ 1 i = neuron with largest net input
a; = i
1 0 otherwise
2 2 , a =w a
0 0
|
1 |
0 |
0 |
0 |
0 |
|
0 |
1 |
0 |
0 |
0 |
|
0 |
0 |
1 |
1 |
1 |
Determine regions of each class and draw a graph to illustrate this.
(b) Write short notes on 4+6
(1) Vector quantization
(2) Unsupervised learning and
its relative merits and limitations with espect to supervised learning.
(c) Discuss Hebb rule in context of : 4+6
(1) Supervised learning
(2) Unsupervised learning.
4 Attempt any two parts :
(a) What do you understand by associative networks ? In this context explain working of a Hopfield network.
(b) Describe brain-in-a-box model. Compare it with a recurrent network.
(c) Derive expression for Boltzmann learning rule. 10
5 Attempt any two parts :
(a) Define gradient. Using steepest descent rule to 10 the following function
determine first three points of trajectory starting from
= [0.5, 0.5]
(b) Discuss the role of 2+4+4
(1) Selection
(2) Cross-over
(3) Mutation
in context of genetic algorithm.
(c) Write an algorithm to implement simulated 10 annealing.
V-1052] 4 [ 960 ]
|
Attachment: |
| Earning: Approval pending. |
