1(a) Draw simple artificial neuron and discuss the calculation of the output. State any few characteristics of any artificial neural network.

1(b) Indicate the difference between excitatory and inhibitory weighted interconnections,

1(c) Compare and contrast BAM and Hopefield network.

1(d) Explain fuzzification and defuzzification process.

1(e) Explain the difference between supervised and unsupervised learning.

2(a) Draw a model of Adaline network. Explain the training algorithm used here.

2(b) What are linearly separable and nonseparable pattern classes? Discuss how perceptrons can be used to classify each of them.

3(a) what are two type of two discrete Hopefield nets? Draw the architecture of discrete Hopefield net. State the testing algorithm used in discrete Hopefield network.

3(b) Draw a simple neural network with a single neuron, four inputs point and one output point.

Apply Hebbian rule to this network wih binary activation function and obtain the updated weight vector. The initial weight vector is

$W^1 = [1 -1 0 0.5]^t$ and the training set consist of three inputs,

$X_1=[1 -2 1.5 0]^t;$    $X_2 = [1 -0.5 -2 -1.5]^t; $    $X_3 = [0 1 -1 1.5]^t $

Assume learning constant as 1.

4(a) What are LVQs? Explain LVQ1 algorithm in detail

4(b) With a neat architecture, explain the training algorithm of Kohonen self organization Feature maps.

5(a) Three fuzzy sets are defined as:

$$A̰= \{ \frac{0.1}{30} + \frac{0.2}{60} + \frac{0.3}{90} + \frac{0.4}{120} \}$$

$$B= \{ \frac{1}{1} + \frac{0.2}{2} + \frac{0.5}{3} + \frac{0.7}{4} + \frac{0.3}{5} + \frac{0}{6} \}$$

$$C= \{ \frac{0.33}{100} + \frac{0.65}{200} + \frac{0.92}{300} + \frac{0.21}{400} \}$$

Find the following:

• (a) $R=A \times B$
• (b) $S=B \times C$
• (c) $T=R o S$ using max-min composition
• (d) $T=R o S$ using max- product composition.

5(b) Explain any four defuzzification methods with suitable diagrams

6(a) Types of activation functions

6(b) Properties of neural networks

6(c) Boltzmann machine

6(d) ANFIS