# CUSAT Test paper Artificial Neural networks Nov 2009

CUSAT B.Tech Degree VII Semester Examination,

November 2009

IT/CS/EC/EI/EB 705(C)

ARTIFICIAL NEURAL NETWORKS

(2002 Scheme)

Time : 3 Hours                                                                              Maximum Marks : 100

1. (a) Explain the working of an artificial neuron.                                             (5)

(b)            What are activation functions? Give an example.                                         (5)

(c)            Explain the working of a multi-layer network with the help of a block diagram.       (10)

OR

1. (a) What do you mean by Linear Separability? Explain a problem that is not linearly separable.               (10)

(b) Explain

(i)             Hebbian learning rule

(ii)            Kohonen learning rule                                                           (10)

1. (a) Explain the architecture of a Back Propagation Network and its working.        (10) (b) Explain the problems involved in Back Propagation Training Algorithms.

(10)

OR

1. (a) Explain what you mean by

(i) Network paralysis                          (ii) Local minima

(iii)            Temporal Instability                                                  (12)

(b) Show how ERROR is propagated back in a Back Propagation Network. (8)

1. (a) Draw the block diagram of a Counter Propagation Network and explain its normal mode of operation. (10)

(b) Explain how weights are initialized in a counter propagation network.            (10)

OR

1. (a) Explain the training algorithm in a counter propagation network.                   (10) (b) Explain any one application of a counter propagation network.                                                        (10)
2. (a) Compare and contrast‘Boltzman Training’and‘Cauchy Training’. (10) (b) What is meant by

(i)             Simulated annealing

(ii)            Artificial specific heat method.                                              (10)

OR

1. (a) What do you mean by Statistical Methods? Explain an algorithm that uses statistical methods.                      (10)

(b) Compare and contrast Back Propagation Training and Cauchy Training.       (10)

1. (a) Draw the architecture of a hopfield network and explain its working. What are the requirements of its weight matrix?                                                                  (10)

(b)         What are genetic algorithms? Explain the mutation and cross over operations.     (10)

OR

1. (a) Draw the architecture of a Bi-directiona! Associative Memory and explain its working. (10) (b) Show how associationare stored and retrieved in a Bi-directional Associative Memory. (10)