Sathyabama University BE CSE/DCS 511702-611701 Neural Networks August 2010

Register Number

                               

SATHYABAMA UNIVERSITY

(Established under section 3 of UGC Act, 1956)

Course & Branch: B.E - CSE/DCS

Title of the Paper: Neural Networks                      Max. Marks: 80

Sub. Code: 511702-611701                                    Time: 3 Hours

Date: 30/08/2010                                                    Session: FN

______________________________________________________________________________________________________________________

                                       PART - A                    (10 x 2 = 20)

                        Answer ALL the Questions

1.     What is Neural Network?

2.     Differentiate biological and artificial neuron?

3.     List down any two applications of perceptrons

4.     Write down the weight updation formula?

5.     Which layer is used for measuring error in back propagation network?

6.     What is Hopfield Network?

7.     How can networks be trained to real world problems?

8.     What is the difference between weight and bias in neural networks?

9.     Differentiate Statistics with Neural approach.

10.   How can image be represented?

PART – B                       (5 x 12 = 60)

Answer All the Questions

11.   Neural Network is not an algorithm but it is a concept-a mechanism. Justify.

(or)

12.   (a) Outline pattern classified mechanism in Neural Networks

        (b) What are the characteristics of Neural Networks?

       

13.   Compare Supervised and Unsupervised learning. 

(or)

14.   Discuss in detail about perceptron training algorithms.

15.   Discuss in detail about Back propagation network.

(or)

16.   System can be modeled and simulated better with counter propagation network. Justify your answer.

17.   With an example discuss in detail about Kohonen network algorithm.

(or)

18.   (a) What is mean field theory?

(b) Can Neural approach be applied for Traveling salesman? Justify your answer.

19.   Outline the steps and implementation mechanism for adaptive resonance theory architecture.

(or)

20.   Consider an image of order M*N pixels. Model and design a Neural Network for the following cases:

        (a) The user wants to manipulate the pixel values and represent it                                                                                              (3)

        (b) For any input pixel value the user wants to represent the closest number of pixels with its value and colour.           (3)

        (c) The user wants to change every white to gray and gray to white color.                                                                               (2)

        (d) Identify and justify the network model designed for the scenario above.                                                                     (4)

Share This