University of Technology
Department of Electrical Engineering
Final Course Examination 2019-2020
Subject: Advanced Intelligent Systems Year: PhD
Division: Electronics & Communications Eng. Time: 2 Hours
Examiner: Prof. Dr. Hanan A. R. Akkar Date: 14 / 6 /2020
Q.1: Choose the correct answer of the following: (Five MCQ only) (20 Marks)
1- A neuron with 4 inputs has the weight vector w = [1 2 3 4]T and a bias Ө = 0 (zero). The activation
function is linear and given by f(net) = (2 net). If the input vector is x = [4 8 5 6]T then the output of
the neuron will be:-
(a) 1 (b) 56 (c) 58 (d) 118
2- If the Genetic Algorithm is applied to minimize the following function:-
F(x) = x2 + 3 where 1 ≤ x ≤ 15
By using single point crossover, roulette wheel selection method and the following initial strings:-
1 1 0 0
1 0 0 1
1 0 1 1
1 1 0 1
The value of x string after applied Genetic Algorithm is:-
(a) 1 1 1 1 (b) 0 0 0 1 (c) 1 0 0 0 (d) 0 0 0 0
3- Artificial Neural Network type MAXNET consists of five inhibitory weights as ( 0.1 ). The net is
initially activated by the input signals (0.7 0.9 0.2 0.8 0.3) with the following activation function:-
x x 0 the wining neuron is:-
x
0 otherwise
a) Neuron 1 (b) Neuron 2 (c) Neuron3 (d) Neuron 5
4- Which of the following statements is the best description of mutation in Genetic Algorithm?
(a) Randomly change a small part of some strings.
(b) Randomly change the fitness function.
(c) Randomly combine the genetic information from 2 strings.
(d) Randomly generate small initial values for the weights.
5- Given a Kohonen Self-Organizing Map with weights as shown in the following diagram, use the square
of the Euclidean distance. The cluster unit CJ that is closest to the input vector (0.5 0.2) is:-
(a) Cluster unit C5 (b) Cluster unit C3
(c) Cluster unit C4 (d) None of these.
6- For the following Artificial Neural Network, if x1=1, x2=2 and x3=1.5 then the output of the network is:-
(a) One (b) Zero (c) 4.5 (d) -1
Page 1 of 2
Note: Answer Two Questions Only of (Q.2, Q.3 and Q.4)
Q.2:A) Calculate the weight matrix for a Hopfield network to store the pattern [ 1 −1 1 −1 ]. (10 Marks)
B) Consider an ART1 neural net with F1 five units and F2 two units. Determine the new weights
matrices after the vectors (1, 1, 1, 0, 0), (0, 0, 0, 1,1), and (1, 0,0,1,1) are represented, where n=5,
1
m=2, L=2, ρ=0.5, bij(0) and tji(0)=1 . (10 Marks)
1 n
Q.3: A) Encode and decode the Hamming network that stores the following patterns:-
X1= [+1.3 -1.4 -0.9 -1] X2= [-1 +0.8 +0.3 -1]
X3= [-1 -0.2 +1.7 +1]
If the following noisy pattern is applied to the Hamming network
X= [+1.2 +0.8 +0.8 -1], Show that X is closest to the patterns X2. (10 Marks)
B) Perform one training step for a single neuron network using the Delta learning rule. Train the
network by using log-sigmoid activation function and the following data pairs:-
3 4
x1 0 , d 1 1 , x 2 2 , d 2 1
1 1
Where, 1 , C= 0.5 and the initial weights are w 1 2 1 1t . (10 Marks)
x2
Q.4: Find the maximum of f ( x) x using genetic algorithm, where:- (20 Marks)
8
The range of x: [0, 1,…,15]
The population size: 4
String length: 4 bits
Crossover type: 1-point crossover (Pc=100%)
Mutation: Bitwise mutation (Pm= 0.001)
Selection method: Roulette wheel selection.
Initialization parents are:-
0 1 1 0
1 0 0 0
0 0 1 1
1 0 1 0
Note: Answer One Question Only (Q.5 or Q.6)
Q.5: A) Compare between the supervised learning rules in terms of the single weight and adjustment
formulas, weight initialization, the neuron activation functions and the requirement for neuron or
layer to proceed the weight adaptation (illustrated in table). (10 Marks)
B) Draw a Flowchart which explains the general structure of the Genetic Algorithm (GA). (10 Marks)
Q.6:A) Compare between GA and ANN. How GA applied in ANN? (10 Marks)
B) 1) What is meant by root mean square error? (10 Marks)
2) What is meant by threshold logic unit?
3) When the training of back propagation network stopped?
مع تمنياتنا لطلبتنا االعزاء بالتفوق والنجاح
االستاذ الدكتوزة حنان عبد السضا
مادة االنظمت الركيت المتقدمت
Page 2 of 2