## Artificial Intelligence - May 17

### Electronics Engineering (Semester 7)

Total marks: 80

Total time: 3 Hours
INSTRUCTIONS

(1) Question 1 is compulsory.

(2) Attempt any **three** from the remaining questions.

(3) Draw neat diagrams wherever necessary.

**Q1) Attempt any four of the following.**

**1(a)**Differentiate between an artificial neural network and digital computer.

**1(b)**What are excitatory and inhibitory weighted interconnections?

**1(c)**Differentiate between supervised and unsupervised learning.

**1(d)**What is a membership function?

**1(e)**Explain the delta rule of learning with an example.

**2(a)**With the help of a flow chart, explain Single Continuous Perceptron Training Algorithm.

**2(b)**Implement the perceptron learning rule for the following set of input training 10 vetors:

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

The learning constant, c=0.1 and the desired responses for $X_1$ , $X_2$ and $X_3$ are $d_1 =-1$, $d_2 =-1$ and $d_3 =1$ respectively. Assume the initial weight vector to be $W_t =[1 -1 0 0.5] ^t$ and obtain the updated weight vector after one epoch.

**3(a)**With the help of a flow chart, explain error back propagation algorithm.

**3(b)**Give the network architecture of an Adaline network and discuss its training procedure.

**4(a)**What are Discrete Hopfield Networks? Explain how patterns are stored in them.

**4(b)**With a neat architecture, explain the training algorithm of Kohonen self-organizing feature maps.

**5(a)**Two fuzzy sets are defined as:

$$A = \{ \frac{1}{2} + \frac{0.3}{4} + \frac{0.5}{6} + \frac{0.2}{8} \} $$

$$B = \{ \frac{1}{2} + \frac{0.3}{4} + \frac{0.5}{6} + \frac{0.2}{8} \} $$

Perform union, intersection, difference and complement over these fuzzy sets.

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

**Q6) Write short notes on any four:**

**6(a)**Learning factors

**6(b)**Perception convergence theorem

**6(c)**Adaptive Resonance Theory

**6(d)**Hebbian learning

**6(e)**Adaptive neuro-fuzzy infonnation system