Theory PDF

Briefly, explain on what you understand about the term ‘machine learning’
in the Artificial Intelligence field.

Solution:
Machine learning, a branch of artificial intelligence, concerns the
construction and study of systems that can learn from data. (Wikipedia)
OR,
Machine learning is programming computers to optimize a performance
criterion using example data or past experience. ( Ethem Alpaydin)
OR,
The goal of machine learning is to develop methods that can automatically
detect patterns in data, and then to use the uncovered patterns to predict
future data or other outcomes of interest. (Kevin P. Murphy)
OR,
Machine learning is about predicting the future, based on the past. ( Hal
Daume III)
Describe TWO benefits of adding ‘intelligence’ in machines.

Solution:
Benefits that can be offered by ‘intelligent systems/machines’ (any two
from the following):
▪ If artificial systems can be made more robust, costly redesigns
can be reduced or eliminated
▪ If higher level of adaptation can be achieved, existing systems
can perform their functions longer and better
▪ If machines can be made to be self-organized then less
operations are needed by humans
Describe the following learning algorithm types and give

ONE related example of techniques for each type.
i) Supervised (inductive) learning
ii) Reinforcement learning
Solution:
Learning algorithms:
- Supervised (inductive) learning: Training data includes desired
outputs/ teacher signals
Examples: Perceptron, Logistic regression, Support vector machine (SVM),
Adaline,
- Reinforcement learning: Given a sequence of examples/states and a
reward after completing that sequence, learn to predict the action to
take in for an individual example/state
Examples: Q-learning, Deep Q-Networks
Describe the 'feedforward' and 'recurrent'
architectures by giving the definition, example of techniques and diagram to
represent each type.
Solution:
Architecture:
- Feedforward: All signals flow in one direction only, i.e. from lower
layers (input) to upper layers (output).
Example: Fuzzy ARTMAP, GRNN, Self-Organizing Maps, Competitive
Learning
- Feedback/recurrent: Signals from neurons in upper layers are fed

back to either its own or to neurons in lower layers.
Example: Adaptive Resonance Theory, Fuzzy ART, Boltzmann Learning,
Hopfield Network
Compare the difference between ‘crisp logic’ and ‘fuzzy logic’ in terms of
precision, membership value and relations.
===================================================
Show typical fuzzy logic controller architecture for a single input-single
output (SISO) feedback system.
• Definitions/explain:
The fuzzification:
involves the conversion of the input/output signals into a
number of fuzzy represented values (fuzzy sets) which comprise of
selection of appropriate Membership Function, appropriate Universes of
Discourse, type of Membership Function and the appropriate center points.
Defuzzification:
is a mapping from a space of fuzzy control actions defined
over an output universe of discourse into a space of non-fuzzy (crisp)
control action
The Inference Engine:

provides the mechanism for invoking or referring to
the rule base such that the appropriate rules are fired. Two most common
methods used in fuzzy logic control are the max-min composition and the
max-(algebraic) product composition.
‘Knowledge Base’ in Fuzzy Logic Control:
- The knowledge base of a fuzzy logic controller consists of a data base and a
rule base.
• The basic function of the data base is to provide the necessary information
for the proper functioning of the fuzzification module, the rule base and the
defuzzification module.
artificial intelligence:
Simulation of human intelligence on a machine such that the machine is efficient enough to identify
and make use of a right piece of knowledge at a given step of solving a problem.
Fuzzy logic:
Biological intelligence: Human reasoning
Function: Produces a reasonable output based on inputs
Application: Servo and regulatory control of a plant
Artificial neural network (ANN):

Biological intelligence: Brain, biological neural network
Function: Learning, classification, identification
Application: Image recognition, prediction based on data

design procedure of a fuzzy logic control system and briefly explain the steps
in the flow chart.
Designing steps:
1. Identify the process input and output variables
2. determine on the number of fuzzy partitions (or fuzzy subsets)
3. Choose the membership functions for the input and output fuzzy
variables.
4. Derive the fuzzy control rules based
5. The inference engine needs to be defined
6. Choose the right choice of defuzzification method
7. Run simulation and check the performance. If the results are not as
desired, tune some parameters.
8. Use trial-and-error approach will continue until satisfactory results
are obtained.
Draw the feedback control block diagram of the water bath system by using
fuzzy logic control approach.
==========================================================
Using the typical practice of fuzzy rule base, complete the fuzzy control
matrix for this system
Illustrate (draw) a biological neuron and an artificial neuron. Identify the
similarities between the two types of neuron.
Similarities: Label axon, cell body and dendrites on both figures

Traditional set theory requires elements to be either part of a set or not. Similarly, binary-
valued logic requires the values of parameters to be either 0 or 1. With a suitable
diagram, argue that with fuzzy sets an element belongs to a set to a certain degree of
certainty.
Fuzzy sets are represented by membership functions that assign each fuzzy set element a truth
value.
Question 5 chapter 5 6 7
• Suggest an ANN training algorithm to be applied. Briefly state its

working principle.
Backpropagation algorithm.
It is based on the gradient descent approach. Weights are adapted to minimize the error
between actual and target outputs.
• Given a set of input and output data, how can the ANN be trained?
• Initialize the ANN

Set learning rate 𝜂 and momentum rate 𝛼
Set minimum error 𝐸𝑚𝑖𝑛
Initialize weights 𝑤𝑗𝑖,𝑤𝑘𝑗 and biases 𝑏𝑖,𝑏𝑗 to random values.
• Calculate total error by applying inputs one at a time and propagate through the layers.
• Back-propagate error through output and hidden layer and adapt weights.
• Back-propagate error through hidden and input layer and adapt weights.
• If 𝐸< 𝐸𝑚𝑖𝑛 stop the training. If not repeat the steps above.
If a new set of input and output data is presented, how can one ensure a
proper alphabet identification by the ANN?
The training is restarted by combining the new data with the old data.

Theory PDF

Uploaded by

Copyright:

Available Formats

Theory PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Theory PDF

Uploaded by

Copyright:

Available Formats

Briefly, explain on what you understand about the term ‘machine learning’

in the Artificial Intelligence field.

Describe TWO benefits of adding ‘intelligence’ in machines.

Describe the following learning algorithm types and give

- Feedback/recurrent: Signals from neurons in upper layers are fed

The Inference Engine:

Biological intelligence: Human reasoning

Function: Produces a reasonable output based on inputs

Application: Servo and regulatory control of a plant

Artificial neural network (ANN):

Function: Learning, classification, identification

Application: Image recognition, prediction based on data

Similarities: Label axon, cell body and dendrites on both figures

• Suggest an ANN training algorithm to be applied. Briefly state its

• Initialize the ANN

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.