Professional Documents
Culture Documents
Soft Computing Aug2021Makeup
Soft Computing Aug2021Makeup
1A) List the important characteristics of Soft computing. Prepare a table to compare soft
computing with hard computing. 5M
1B) What is expert system? With a neat block diagram architecture of expert system explain each
of its block. Also elaborate your explanation on functioning of expert system when these blocks
are integrated. 5M
2A) Draw the Gaussian Membership Function distribution graph for Gaussian function. Find out
the membership function value corresponding to the given input value x = 9.0 with mean ( c ) =
10.0 and standard deviation ( σ ) = 3.0 Also determine the α-cut and strong α-cut for α = 0.6 4M
2B) The fuzzy sets P and Q are defined over the universal set X
P = { (x1, 0.1), (x2, 0.2), (x3, 0.7), (x4, 0.5), (x5, 0.4)}
Q = { (x1, 0.9), (x2, 0.6), (x3, 0.3), (x4, 0.2), (x5, 0.8)}
Find the crisp set for the following:
i) P0.2 and P0.3 ii) (P ∪ Q)0.6 iii) (P ∪ ¬P)0.8 iv) (P ∩ Q)0.4
using Lambda-cut method. 6M
3A) Design a neural network that satisfies the truth table given in Fig.Q3A using McCulloch Pitt
model principles. x1 and x2 are the two inputs and Y is the output of the neural network.
Fig. Q3A
Extend your design further to complement the output Y.
Draw the diagram of such a neural network with the designed values. Firing of the neuron is based
on usage of the following activation function:
f(z) = 1, if z ≥ θ
f(z) = 0, if z < θ 5M
Page 1 of 3
3B) Write down the equations and related derivatives (Refer Fig. Q3B) of W19, W16 and W1
using Chain-rule to update the weights during Back propagation. The deduced derivatives in your
equations must be ready to substitute the numeric values directly: 5M
Fig. Q3B
4A) Train an Auto-associative neural network for input vector [-1 1 1 1] and also test the
network for the same input vector. Test the neural network separately for one mistake entry and
two missing entries in the input vector. Give your conclusion on the result of all the tests carried
on. Draw a neat diagram of Auto-associative neural network by labelling all the design parameters.
5M
4B) Construct a Kohonen’s neural network to cluster three given vectors [1 1 0 0], [0 0 0
1], and [1 0 0 0]. It is required to form two clusters. The initial synaptic weights in the Kohonen’s
neural network are given to be W11 = 0.2, W12 = 0.8, W21 = 0.6, W22 = 0.4, W31 = 0.5, W32 = 0.7,
W41 = 0.9 and W42 = 0.3 where Wij is the synaptic weight of ith neuron of the input layer to jth
neuron of the output layer in the Kohonen’s neural network. The learning rate is given to be 0.6
Write your conclusion after analyzing each of the four given vectors. Draw a neat diagram of
Kohonen’s neural network by labelling all the design parameters. 5M
5A) Draw the flow chart of Simple Genetic Algorithm. Discuss in detail the working principle of
it. 5M
5B) Consider a real variable with bounds [8, 35] to be represented by a 3-bit binary string. Prepare
a table for how to convert the 3-bit binary string from decoded value to real value. What is the
obtainable accuracy ϵ. From the values you obtained above, represent a chromosome with any of
the three of the real values to the Genes in phenotype. 5M
6A) The Rule base to be followed for a Neuro Fuzzy system is given in Fig. Q6A where in I1 and
I2 are the inputs and O is the output of the Controller. The neural network will consist of five
layers. The input I1 has been expressed using two linguistic terms: Low (LW) and High (H).
Similarly the input I2 has been expressed using 3 linguistic terms: Very Near (VN), Near (NR) and
Far (FR).
The output O has been expressed using 3 linguistic terms: Slow (S), Fast (F). Draw a related neural
network that assist to design a Fuzzy logic controller for a Neuro Fuzzy system. 5M
Page 2 of 3
Fig. Q6A
6B) For two of the fired rules in Neuro-Fuzzy system, the graphical representation for fuzzified
output is shown in Fig. Q6B. Calculate the area and center of area of each of them and finally
compute the crisp value of combined fuzzified output using Center of Sum (CoS) method. 5M
Fig. Q6B
Page 3 of 3