B.Tech-B.Tech Biomedical Engineering Artificial Neural Networks(Sathyabama University, Chennai, Tamil Nadu-2010)
Friday, 23 August 2013 03:50Duraimani
(Established under section 3 of UGC Act,1956)
Course & Branch :B.Tech - BME
Title of the Paper :Artificial Neural Networks Max. Marks :80
Sub. Code :524606/BME16 Time : 3 Hours
Date :03/05/2010 Session :FN
PART - A (10 x 2 = 20)
Answer ALL the Questions
1. What are the computational advantages offered by artificial neural networks?
2. Give the drawbacks of Hebbian Learning.
3. What is the need for a nonlinear activation function in a multi layer neural network?
4. Why should we train a neural network?
5. Write the equation for bipolar sigmoidal activation function.
6. What is meant by attractors in a dynamical system?
7. Draw the block diagram of a Hetero Associative Memory.
8. Define orthonormality for any neural network.
9. State any two applications of ART network in image processing.
10. List the characteristics of Energy function in Hopfield networks.
PART – B (5 x 12 = 60)
Answer All the Questions
11. Perform the perceptron rule training of the single neuron network using the training pairs.
and the intial weight of the network is w = [3, 2, 6, 1]t. Assume the learning constant to be 1.
(or)
12. Explain the structural organization of Human brain with neat sketch.
13. The weight matrix W for a network with bipolar discrete binary neurons is given as
W = W-1. Knowing that the threshold 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
(or)
14. Derive the expression for the Back – propagation training algorithm.
15. Explain the applications of Cauchy training algorithm to the general non-linear optimization problems.
(or)
16. Explain the artificial specific heat method with neat architecture.
17. A Hopfield neural network required to store the two fundamental memory patterns given by V1 = [1 1 1 -1] and V2 = [-1 -1 -1 1].
(a) Find the 4-by-4 synaptic weight matrix of the network.
(b) Draw the architecture of the Hopfield network.
(c) Compute the values of energy for the given input patterns V1 and V2
(d) Why is only asynchronous updation used in Hopfield networks?
(or)
18. Explain the Traveling salesman problem with one example following the constraints.
19. A three bit auto associative recurrent memory updating under asynchronous rule stores a single vector s(1) = [1 1 -1]t. Prepare a state transition map using a three dimensional cube with the energy values assigned to each vertex, and with the direction of all possible transitions assigned to each of the cube. Then evaluate all transitions using each vertex as an initial condition.
(or)
20. Explain in brief
(a) Adaptive resonance theory (b) ART 1 and ART 2.
| Earning: ₹ 3.25/- |
