You are on page 1of 5

Code No: RR410212 Set No.

1
IV B.Tech I Semester Supplimentary Examinations, February 2008
NEURAL NETWORKS AND APPLICATIONS
(Electrical & Electronic Engineering)
Time: 3 hours Max Marks: 80
Answer any FIVE Questions
All Questions carry equal marks
⋆⋆⋆⋆⋆

1. (a) With the help of a neat diagram explain the analogy of a logical neuron.
(b) Explain what is an artificial neural network and show how a basic ANN is
constructed using a biological neuron. [8+8]

2. (a) Distinguish between local minima and global minima in neural networks and
what are the effects of these on neural networks.
(b) Explain the distinction between stability and convergence. [8+8]

3. Write brief note on:

(a) Pattern
(b) Classes
(c) Pattern space
(d) Decision regions. [16]

4. Derive the learning rule for Back Propagation network. What are the major draw-
backs. Suggest solutions, to over come these drawbacks. [16]

5. (a) Define recurrent network. Give some examples.


(b) Draw the flowchart of producing solution of optimization problems using feed-
back networks. [8+8]

6. What are the applications of feature mapping algorithm? Explain taking any one
of the application as example. [16]

7. Derive a numerical solution for finding the solution of differential equation. [16]

8. Consider the following neuron circuit with three weights. Following as the data on
transistors figure 8, control voltages and the circuit elements used to design the
circuit transistors: k1 = 20µA/v2 vth = 1v
L L L
  1

W 1
= 1, W 2
= 3
, W 3
= 13
control voltages vgs1 =5v, vgs2 =3v, vgs3 = 3.75v circuit RF = 100 kΩ.
Find the neuron’s output voltage v0 as a function X where X=[x1 x2 x3 ]T . [16]

1 of 2
Code No: RR410212 Set No. 1

Figure 8

⋆⋆⋆⋆⋆

2 of 2
Code No: RR410212 Set No. 2
IV B.Tech I Semester Supplimentary Examinations, February 2008
NEURAL NETWORKS AND APPLICATIONS
(Electrical & Electronic Engineering)
Time: 3 hours Max Marks: 80
Answer any FIVE Questions
All Questions carry equal marks
⋆⋆⋆⋆⋆

1. (a) Describe the characteristics of artificial neural networks.


(b) Compare Artificial Neural Networks with Biological Neural Network. [8+8]

2. (a) What are the requirements of learning laws.


(b) Distinguish between activation and synaptic dynamics models. [16]

3. Explain perceptron training for an image recognition system. [16]

4. Design and train a feed forward networks for the problems.

(a) Consider a 4 input and 1 output problem where the Output required to be
‘one’, if the input configuration is symmetrical and ‘zero’ otherwise.
(b) Why back propagation is also called as generalized delta rule. [12+4]

5. (a) What are the advantages of vector field method over other methods?
(b) The weight matrix W for a network with bipolar discrete binary neurons as
given as:
 
0 1 −1 −1 −3
 1 0 1 1 −1 
  −1
W =  −1 1
 0 3 1 Ω
 −1 1 3 0 1 
−3 −1 1 1 0
knowing that the thresholds and external inputs of neurons are zero, compute
the values of energy for v = [-1 1 1 1 1]t and v = [-1 -1 1 -1 -1]t . [4+12]

6. Write a short notes on:

(a) Hamming NET


(b) MAXNET. [8+8]

7. State the linear programming problem. Explain how to solve optimization problems
using neural networks. [16]

8. Explain template matching networks in neural processing. Draw a template bit


map and the corresponding circuit diagram. [16]

⋆⋆⋆⋆⋆

1 of 1
Code No: RR410212 Set No. 3
IV B.Tech I Semester Supplimentary Examinations, February 2008
NEURAL NETWORKS AND APPLICATIONS
(Electrical & Electronic Engineering)
Time: 3 hours Max Marks: 80
Answer any FIVE Questions
All Questions carry equal marks
⋆⋆⋆⋆⋆

1. (a) Differentiate between supervised and unsupervised learning.


(b) Explain how instar learning rule comes under supervised learning rule and
outstar learning comes under supervise learning rule. [8+8]

2. (a) Distinguish between local minima and global minima in neural networks and
what are the effects of these on neural networks.
(b) Explain the distinction between stability and convergence. [8+8]

3. (a) What is meant by perceptron representation problem.


(b) Explain why a single layer perceptron cannot be used to solve linearly non-
separable problems. Give two examples for linearly non-separable problems.
[16]

4. The linearly separable patterns X1 − − − X10 listed below have to be classified


in two categories using a layered network and an appropriate pattern image space
transformation. Design a two layer classifier with the bipolar discrete perceptrons
based on appropriate space mapping.
           
1 3 1 2 3 2
X1 = X2 = X3 = X4 = X5 = X6 =
3 3 2 2 2 1.5
 
5
[X7 ]T = −2 0 [X8 ]T = 1 0 [X9 ]T = 3 0 [X10 ] =
     
= [X10 ]
0
X4, X6, X7, X8, X9: Class 1; remaining pattern Class 2. [16]

5. Construct energy function for a discrete and continuous Hopfield neural network,
each of size N×N neurons. Show that the energy function decreases everytime the
neuron output has changed. [16]

6. Differentiate between Hamming NET and MAXNET. [16]

7. Derive a numerical solution for finding the solution of differential equation. [16]

8. Explain how neuro computing circuits can be modeled using digital and analog
circuits. [16]

⋆⋆⋆⋆⋆

1 of 1
Code No: RR410212 Set No. 4
IV B.Tech I Semester Supplimentary Examinations, February 2008
NEURAL NETWORKS AND APPLICATIONS
(Electrical & Electronic Engineering)
Time: 3 hours Max Marks: 80
Answer any FIVE Questions
All Questions carry equal marks
⋆⋆⋆⋆⋆

1. (a) Show that the sigmoidal function.


s(x) = 1/ (1+e−cx ) has an activation derivative
s’(x)= cs(1-s)
(b) Compare the machine and brain learning. [8+8]

2. (a) What are the requirements of learning laws.


(b) Distinguish between activation and synaptic dynamics models. [16]

3. Explain how to interpret perceptron learning as a gradient descent algorithm. What


is the gradient term here? [16]

4. (a) Distinguish between Multi-layer Perceptron and Multi-layer feed forward neural
network.
(b) What are the ill-possed problems in the context of training a Multi-layer feed
forward network. [16]

5. The famous Eight Queens problem is the one where eight queens are to be placed on
an 8×8 chessboard so that no one queen can take another. Construct the Energy
function that needs to be optimized to solve the Eight Queens problem. From
this energy function, derive the thresholds of nodes and the weight matrix of the
Hopfield network that can be set up to solve this problem. [16]

6. Mention some of the feature mapping capabilities of neural networks and explain
any two of them in detail. [16]

7. Derive a numerical solution for finding the solution of differential equation. [16]

8. Explain how neuro computing circuits can be modeled using digital and analog
circuits. [16]

⋆⋆⋆⋆⋆

1 of 1

You might also like