NEURAL NETWORKS : B.E. BIOMEDICAL ENGINEERING

[Kalyan]

NEURAL NETWORKS

B.E. BIOMEDICAL ENGINEERING

6TH SEM

NEURAL NETWORKS

MAX:100
ANSWER THE FOLLOWING:


PART A-(10*2=20)
1. Draw the stimulated structure of a neuron.

2. What is meant by learning rate parameter?

3. How weights are initialized by BAM?

4. Mention the special features of Boltzman machine.

5. Draw any two activation functions.

6. What is an instar?

7. Compare ART 1 and ART 2 (any 2)

8. State the significance of Mexican hat function in SOM.

9. Define spatiotemporal pattern

10. What are 'S' cells?
PART-B (5*16 -80)

11. (a) (i) Describe the biological neuron and brief the feature of artificial neural networks.

(ii) Solve the EXOR problem with perceptron.

(Or)

(b) Explain the madaline architecture and describe the MR II learning algorithm. How madaline is used for translation invariant pattern recognition?

12 (a) (i) Describe the learning expressions in the back propagation network.

(ii) Describe the generalized delta rule.

(Or)

(b) Describe the structure and operation of continuous Hopfield network. & Construct an autoassociative BAM using the following training vectors. X1 = (1,-1,-1,1,-1,1)t and x2 = (1,1,1,-1,-1,-1)t . Determine the output using xo =(1,1,1,1,-1,1)t

13 (a) Describe the concept of simulated annealing. How this is applied in Boltzman machine to overcome the drawbacks of Hopfield network. Also ngive the procedure for retrieving the stored vector with partial knowledge in Boltzmann machine.

. (Or)

(b) Describe the architecture and training process of Counter propogation network in detail.

14(a) What are the special features of ART networks? How pattern matching is done in ART1?

(Or)

(b) Explain the SOM learning algorithm in detail. How will you perform the ballistic movement of a robot arm using SOM?

15(a) What is a neocognitron? With suitable diagram explain the cell S and C cell processing algorithm of neocognitron.

(Or)

(b) Discuss how spatio network is used for speech recognition.