Name : …………………………………………….……………… Roll No. : …………………………………………...…………….. Invigilator’s Signature : ………………………………………..
CS/M.TECH (CSE)/SEM-2/CSEM-205A/2011
2011 SOFT COMPUTING
Time Allotted : 3 Hours
Full Marks : 70
The figures in the margin indicate full marks.
Candidates are required to give their answers in their own words
pap
as far as practicable.
Answer any five of the following :
a)
wb er.
1.
5 × 2 = 10
What is the difference between a threshold function and a sigmoid function ?
b)
What is the stochastic model of a neuron ?
c)
What is the role of bias in a non-linear model of a
d)
Give two reasons for introducing hidden units in a multilayer perceptron.
e)
in a c. ut .
neuron ?
What is the use of cost function in optimization techniques ?
30410 (M.Tech)
[ Turn over
CS/M.TECH (CSE)/SEM-2/CSEM-205A/2011
Explain the idea of iterative descent in optimization
/q p:/ htt
f)
algorithms.
g)
What are the features and rules in biological systems
those are found in GA-based softwares ?
2.
4 × 5 = 20
Answer all the following questions :
a)
Show
that
the
following
function
satisfies
the
pap
requirements of a sigmoid function.
wb er.
b)
Describe briefly competitive learning in the context of artificial neural network.
c)
Explain the differences between Delta rule and Hebb's rule.
What is the training problem for the elementary perceptron ?
3.
in a c. ut .
d)
Answer any two of the following questions. a)
2 × 20 = 40
Deduce the perceptron convergence algorithm when the learning-rate parameter is of unit magnitude.
30410 (M.Tech)
2
CS/M.TECH (CSE)/SEM-2/CSEM-205A/2011
The training of a multilayer perceptron is done by back-
/q p:/ htt
b)
propagation learning algorithm. Deduce the expression for the local gradient when the neuron is an output node.
c)
Consider the fitness function f(x) = xn on the interval
x € [0, 1]. Calculate the expected population average of a randomly selected population of points.
wb er.
pap 3
in a c. ut .
30410 (M.Tech)
[ Turn over
CSEM-205A _Final_.pdf
f) Explain the idea of iterative descent in optimization algorithms. g) What are the features and rules in biological systems those are found in GA-based softwares ...
