UEC704
UEC704
la. Realize a ANN based on Hebbs learning law for: (i) AND gate (ii) XOR gate (5)
Assume bipolar inputs and targets. Initial weights and biases are '0'. Comment on the results and justify.
b. What are the limitations of Back-prop Learning? (2)
c. Write the expression for Energy of discrete Hopfield Network. Show that energy of discrete Hopfield
Network always decreases. (3)
2 a. Develop MP, Perceptron and Adaline ANN model of 'OR' gate. (5)
b. For a KSOM with 2 cluster units and 3 input units. The weight matrix is [0.9 0.4;0.7 0.3; 0.6 0.5].
Assuming Learning rate=0.2 and Euclidean distance as discriminant function. Perform one epoch of
learning and find updated weights, when input applied is [0.4 0.2 0.1] (5)
3a. Explain different classes of problem complexity based on Time. (6)
b. Differentiate between hard Computing and soft Computing. (4)
0.6
0.3
Find the defuzzified output using (i) centroid method (ii) Weighted Average method (5)
b. If X=[0, 20, 40 60 ,80,100] and three fuzzy sets defined over X are given as:
A={0.5/20, 0.65/40,0.85/60, 1/80,1/100} B=(0.35/20,0.5/40,0.75,60,0.9/80,1/100} (5)
(i) Verify the De-Morgans property
(ii) Define resolution identity.
(iii) Compute difference operation set C=A-B and find lambda cut set on set C corresponding to a=0.3
6. Solve the following problem using f(x)= x2 +3 as maximization problem using Genetic algorithm. Use
rank selection, uniform crossover (mask-11001)and mutation with Pm=0.1 to find the optimal solution
after one epoch of simple GA. Assume 5 bit binary encoding. Assume population size of 5. (10)
7a. For a chromosome with eight elements which will give higher accuracy: Octal or Binary . Explain (2)
b. Assume a tour of 8 cities for TSP. For Parents p1= 1 3 5 6 4 2 8 7 and p2= 1 4 2 3 6 5 7 8 find the
offsprings generated by
(i)Partially match crossover (ii) cyclic cross over.
(4)
c. Also perform different types of mutation used for TSP .
(4)