B.E-B.E Computer Engineering NEURAL NETWORKS AND FUZZY SYSTEMS(Karunya University, Coimbatore-2012)
Saturday, 24 August 2013 12:47anudouglas
Karunya University
(Karunya Institute of Technology and Sciences)
(Declared as Deemed to be University under Sec.3 of the UGC Act, 1956)
End Semester Examination – November/ December - 2012
Subject Title: NEURAL NETWORKS AND FUZZY SYSTEMS Time: 3 hours
Subject Code: 09EC213 Maximum Marks: 100
Answer ALL questions
PART – A (10 x 1 = 10 MARKS)1. Name three components in biological neuron.
2. Draw the characteristics of hyperbolic tangent activation function.
3. Name one of the unsupervised learning rules.
4. Name two different types of memory network for pattern association.
5. What is self-organization in a neural network?
6. Give an example of a recurrent network.
7. Define the term ‘Fuzzy’?
8. Name any two membership functions.
9. What is the Defuzzification method used in fuzzy classification by equivalence relations?
10. What is meant by Pattern recognition?
PART – B (5 x 3 = 15 MARKS)
11. Name the different types neural network architecture.
12. Draw the Hopfield neural network model.
13. Define activation function and weight matrix used in MAXNET.
14. What are the conditions to be satisfied for a relation to be an equivalence relation?
15. What is the difference between classification and pattern recognition?
PART – C (5 x 15 = 75 MARKS)
16. a. Explain the various elements of artificial neural network with functionalities. (7)
b. Write the mathematical equation of a commonly used activation function along with its characteristics. (8)
(OR)
17. With a neat diagram explain the architecture and training algorithm of the Back propagation network.
18. Find out the weights of the Perceptron network with AND function at the end of second iteration for the following condition and comment on the results.
i. Assume input is binary and bias input of 1. α=1, bias weight and initial weights are zero. Output is bipolar.
ii. Assume input is bipolar and bias input of 1. α =1, bias weight and initial weights are zero. Output is bipolar.
(OR)
19. Design a Hopfield network to store the pattern X=[1 1 1 -1]. Test the performance of the network with the inputs [1 1 1 1], [1 1 -1 -1], [1 -1 1 -1] and [-1 1 1 -1].
[P.T.O]
20. Design a Hamming network to store exemplar vectors [1-1 -1 -1] and [-1 -1- 1 1]. Test the performance of following input patterns in the Hamming net to find the closest exemplar.
(i) [1 1 -1 -1] (ii) [1 -1 -1 -1] (iii) [-1 -1 -1 1] (iv) [-1 -1 1 1]
(OR)
21. Design an ART1 network to cluster the following vectors.
[1 1 0 0] and [0 0 0 1]
Assume maximum number of clusters to be formed is 3. Vigilance parameter =0.4. Initial top-down weights =1. Assume parameter to update bottom-up weight is (L)=2.
22. The membership function fuzzy sets of representing resistance (Re), current (I) and speed (N) of DC motor is given below.
;
Find the relationship R= Re X I and S=I X N. Also find Max-Min composition of R°S.
(OR)
23. a. Brief about various features of membership functions. (5)
b. Explain different methods of defuzzification. (10)
24. Explain in detail any two methods of Fuzzy classification.
(OR)
25. Explain in detail the design of a fuzzy controller to maintain the temperature of water heater. Assume inputs to the controller are temperature of water varying from 00C to 1250C and level of water in the heater varying from 0 to 10. Output of the controller is the knob position varying from 0 to 10 to adjust the flow of steam to maintain the temperature of water at 600C.
| Earning: ₹ 5.75/- |
