You are on page 1of 5

1) draw and label component of fuzzy logic controller

i) fuzzification to convert numerical values to linguistic values


ii) rules evaluation which includes a) fuzzy inference - process of mapping from a given input to
an output using the theory of fuzzy sets and b) aggregatation unification of all outputs
iii) defuzzification to convert the linguistic results to numerical values

2) 3 example of tuning technique

When the result is not as expected, then we have few options to change:
1. increase/decrease the number of linguistic values, then we need to adjust the rules again and test
OR
2. change the membership functions for any of the linguistic values, and then check the results
3. adjust/add more rules
This has to be done until the program is successful. Now, sometimes simulation is not possible. Thus,
tuning using real system is really costly.


3) Design fuzzy system for washing machine
i) define i/o: input = weight, dirtiness, output = washing time
[1 mark]
ii) define fuzzy sets for each variables
weight = light, medium, heavy
dirtiness = mild, medium, dirty
washing time = short, medium, long
[3 mark]
iii) define membership function for each fuzzy sets

min and max for every membership function
unit for every input and output axis





[6 mark]
iv) define units and scale
[1 mark]
v) define rules
if weight is medium and dirtiness is mild then washing time is medium
if weight is light and dirtiness is dirty then washing time is long
if weight is heavy and dirtiness is medium then washing time is long

[3 mark]
vi) fuzzification
convert real scalar into fuzzy value via determined fuzzifier and fuzzy rules.
implement fuzzy AND operator

viii) defuzzification
converting fuzzy output scalar (degree of membership) into real scalar value
Sugeno Fuzzy Interference method used to have a single spike. The spike is unity at a single
particular point on the universe of discourse and zero for everywhere. Final step is to determine
how much the volume should be chosen. Weighted Average (WA) formula use to get the crisp
output.
k = [((volume 2 - volume 1) x dom) + (volume 1)] x dom
Weighted Average (WA) = [k(less) x dom(less)] + [k(normal) x dom(normal)] + [k(more) x dom(more)]
k(less) + k(normal) + k(more)

4) Analogy between biological and artificial neural network :




Biological Neural network Artificial Neural network

Soma (Cell Body) Neuron
Dendrite Input
Axon Output
Synapse Weight


5) Draw and label a multi-layer neural network with two input neurons, three hidden
neurons and two output neurons.
[5 marks]

a) Based on the MLP neural network drawn in a) above, find the output at each neuron in
the hidden and output layers if the following parameters are used:
x1=1, x2= -1, all thresholds = 0.1, all weights = 1, all activation functions using sign.
[15 marks]


(

=

=
n
i
i i
w x sign Y
1
u
y3 = sign[x1w13+x2w23-] = sign[1+(-1)-0.1] = sign[-0.1] = -1
y4 = sign[x1w14+x2w24-] = sign[1+(-1)-0.1] = sign[-0.1] = -1
y5 = sign[x1w15+x2w25-] = sign[1+(-1)-0.1] = sign[-0.1] = -1
y6 = sign[y3w36+ y4w46+y5w56-] = sign[-1+(-1)+(-1)-0.1] = sign[-
3.1] = -1
y7 = sign[y3w37+ y4w47+y5w57-] = sign[-1+(-1)+(-1)-0.1] = sign[-
3.1] = -1
[3 marks each]

[2 marks for label]
[3 marks for neurons and connection]

1
2
6
7
4
5
3
Input Hidden Output
6) Describe the adaptive learning rate used in training algorithm in a multilayer neural
network.
[5 marks]



The small learning rate parameter, , causes small changes to the weights in the
network from one iteration to the next, and thus leads to the smooth learning
curve.
On the other hand, if the learning rate parameter, , is made larger to speed up the
training process, the resulting larger changes in the weights may cause instability
and, as a result, the network may become oscillatory.
[1 mark]

To accelerate the convergence and yet avoid the danger of instability, the following
heuristics are applied:
- Heuristic 1. If the change of the sum of squared errors has the same algebraic
sign for several consequent epochs, then the learning rate parameter, should
be increased.
- Heuristic 2. If the algebraic sign of the change of the sum of squared errors
alternates for several consequent epochs, then the learning rate parameter,
should be decreased.
[4 marks]

You might also like