Trends in Computer Science

4COSC0006C

BEng (Hons) Software Engineering

Machine Learning
Overview of machine learning.
Comparison between machine learning and conventional computing.

Student Name: Vihangi Patterson

UOW ID: W19530646
Student ID: 20220011
DATE: 28.10.2022

0
Table of Content

Introduction..………………….……………………………………..
…………………………………..….3

1. Machine
Learning…………………………………………………….............................
.......4

1.1 Types of Machine

Learning……………………………………………………………….
…….5

1.1.1 Supervised
learning…………………………………………………………….
………5

1.1.1.1 Classification
…………………………………………………….……6

1.1.1.2 Regression
………………………………………………….
………….7

1.1.2 Unsupervised
learning………………………………………………………….
……..8

1.1.3 Reinforcement
learning…………………………………………………………….
…9

2. Conventional Computing ………………...

……………………………………………………….10

3. Comparison Between Machine learning and Conventional

computing……..11

Conclusion………………………………………………………………………
……………………………...12
1
Reference…………………………………………………………………………
……………………………..13

Table of figures

Figure 1 Types of machine learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))________________5
Figure 2 Supervised learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))______________________5
Figure 3 Supervised learning: classification (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))__________6
Figure 4 Unsupervised learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))____________________8
Figure 5 Reinforcement learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))___________________9
Figure 6 Conventional computing (Author created)________________________________________________10

2
Introduction

The most recent chess world champion was defeated more than 20 years ago by a computer
program. IBM's Deep Blue computer program defeated world chess champion Gary
Kasparov. There was probably a lot of interest in machine learning at that time.

Machine learning is a widely used technology that is used in almost every aspect of life.
Google has become one of the leaders in the field because to a significant portion of its
research being focused on machine learning and artificial intelligence. Google's two most
difficult projects in its pursuit of machine learning innovations are Google Brain and Google
Self-Driving Car. In conclusion, machine learning has developed into a way of life regardless
of what aspect of life we closely investigate. But where did it all start?

Simply said, everything started in the past. Conventional computer programming and
computation pave the way for revolutionary machine learning. The disadvantages and
weaknesses in traditional computing has encouraged man kind to improve and develop the
technology. Since its conventional computing that has been make machine learning possible
the comparison between the two is not very complicated.

1. Machine learning
3
There are numerous ways to define machine learning, but Tom Mitchell, a professor in the
school of computer science at Carnegie Mellon University, offers the following definition.
computer learning is,

If a computer program's performance at tasks in a class of tasks T, as assessed by a

performance measure P, increases with experience E, it is said to have learned from
experience E with respect to that class of tasks T and performance measure P.

The key point from this is that a machine can be called to learn if it is able to accumulate
experience while completing a certain task and then improve its performance while carrying
out comparable tasks in the future. When discussing a task, the term "past experience"
refers to earlier knowledge on the subject. This data gives the gadget a source of input. And
it is essentially a machine learning.

Examples:

 E denotes the enjoyment of the game, T the work involved in playing checkers, and P
the performance metric, which is the player's victory percentage.

 The same mapping can be used to resolve any other machine learning problem, such
as a problem with image categorization.
 In the context of image classification, E stands for historical data containing images
with labels or that have been assigned classes.

The three main areas of research that make up machine learning are as follows:
 task-oriented research
 cognitive exercise
 theoretical examination

Applications of machine learning:

 Computer Programming
 Game playing
 Agriculture
 Music
 Mathematics
 Medical diagnosis

4
1.1. Types of machine learning

Three categories serve as a foundation for machine learning.

Figure 1 Types of machine learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))

1.1.1 Supervised learning

 In order to train a function that transforms an input into an output,

supervised learning often uses sample input-output pairs. Various training
methods and labeled training data are used to find function instances.

Figure 2 Supervised learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))

Examples for supervised learning:

5
 the ability to foresee a game's outcome

 determining the malignancy or benignity of a tumor

 estimating the cost of several domains, including real estate,

stocks, etc.

 Putting texts into categories, for example designating a group

of emails as spam or not spam

1.1.1.1 Classification

o The issue is referred to as a classification challenge when we're

attempting to predict a nominal or categorical variable.

o The core issue is how to categorize or classify test data based on the
training data's information about labels, categories, and classes. The
computer must map a new image or test data to a set of images to
which it is comparable and assign the same label or category to each
image.

Figure 3 Supervised learning: classification (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))

1.1.1.2 Regression
6
 In the case of a linear regression, the least squares approach is used to "fit"
a straight-line relationship between the predictor variables and the targets.
A straightforward strategy would be to use a simple linear regression model
with investment as the predictor variable and sales revenue as the goal
variable. A straightforward strategy would be to use a simple linear
regression model with investment as the predictor variable and sales
revenue as the goal variable.

The typical applications of regression are shown in :

o projecting sales for managers using retail demand

o Weather forecasts for estimating real estate prices
o Forecast of skill demand in the labor market

Regression's typical uses can be observed in

 retail demand forecasting

 predicting sales for managers
 Real estate price estimation
 weather predictions
 Job market skill demand forecast

1.1.2 Unsupervised learning

7
 Unsupervised learning is one example of a data-driven system that analyzes
unlabeled data without a human's help. This is widely used for experimental
reasons, groupings of results, identifying pertinent trends and structures,
and extracting generative features.

Figure 4 Unsupervised learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))

Examples for unsupervised learning

 K-Means Clustering,
 Principal Component Analysis
 Hierarchical Clustering.

8
1.1.3 Reinforcement learning

 Reinforcement learning is a potent technique for developing AI

models that can raise productivity or increase automation in complex
systems like manufacturing, logistics in the supply chain, autonomous
cars, and robots. Computers and software agents are capable of
automatically analyzing the most effective conduct in a given situation
or environment.

Figure 5 Reinforcement learning (Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018))

Examples for reinforcement learning:

 Self-driving automobiles are one modern instance of
reinforcement learning. The crucial data that it must
monitor includes the speed and speed limits on various
sections of road, as well as traffic, road, and weather
conditions. Start and stop, acceleration and deceleration,
turning to the left or right, and other operations must be
performed.

9
2. Conventional computing

A traditional phenomenon of computing devices is conventional computing. Conventional

computers primarily do two tasks: first, they store data or information in memory, and
second, they process that data using calculation, algorithms, and formulae to provide results
based on the following requirements:

 based on the concept of electrical circuits, which have two states: ON and OFF.
 A bit can be in one of two states, either 0 or 1.
 Circuits are used to carry out the calculations.
 All these gadgets can only be in one of two states: On or Off.
 Metal oxide semiconductors are used to create the CMOS transistors, which are the
ones that are most frequently used.

Traditional computing is always accompanied with conventional programming.

In traditional programming, a defined set of scenarios are fed to the computer along with a
set of instructions. After then, a machine will use its computing power to assist a human in
quickly and effectively processing the data.

Data

Computer Output
Program

Figure 6 Conventional computing (Author created)

10
3. Comparison of machine learning and conventional computing

Conventional computing Machine learning

Manual process Automated process

Write down the exact steps required to solve Motivated by human behavior and solve the
a problem problems according to examples

Accuracy of predictions are low Accuracy of predictions are high

Take more time to resolve problem Take less time to resolve a problem

based on the concept of electrical circuits a subset of artificial intelligence

Human intervention is needed No need of human intervention

Table 1

11
Conclusion

Traditional computing, which uses a manually generated program that runs on a computer
and uses input data to produce output, has been around for a while.

Machine learning, which automates the process of algorithms creating rules from data, is an
effective method for producing insightful data that may later be used to forecast profitable
business outcomes.

In my view, conventional computing as well as programming will not replace by machine

learning (ML). Because the purposes and usability of each varies. For the development and
construction of websites, Android and iOS apps, and other things, conventional
programming will always be used. However, machine learning (ML) can be utilized to create
intelligent systems like fraud detection, recommendations, and more.

12
References
1. Dutt,A. and Chandramouli,S. and Kumar Das,A.(2018).Machine learning: Machine
learning. Available from https://learning.oreilly.com/library/view/machine-
learning/9789389588132/ [Accessed 19 October 2022].

2. Sarkar,I.H. (2021). Machine Learning: Algorithms:Real-World Applications and

Research Directions. SN computer science,2(3),1-21.Available from
https://link.springer.com/article/10.1007/s42979-021-00592-x [Accessed 17
October 2022]

3. Alzubi,J. and Nayyar,A and Kumar,A.(2018). Machine Learning from Theory to

Algorithms: An Overview. Journal of physics: conference series,1142.Available

from https://iopscience.iop.org/article/10.1088/1742-6596/1142/1/012012/
pdf [accessed 17 2022]

4. Seife,C.(2005).What Are the limits of Conventional Computing. Science

(American Association for the Advancement of Science),309 (5731), 96-96.
Available from https://www.science.org/doi/pdf/10.1126/science.309.5731.96
[Accessed 21 October 2022].

5. Bhatia,R.(2018).How do machine learning differ from Traditional algorithms.

How do machine learning differ from Traditional algorithms. Available from
https://analyticsindiamag.com/how-do-machine-learning-algorithms-differ-
from-traditional-algorithms/ [Accessed 22 October 2022]

6. Faroukh,Y.M.(2018). Quantum Computers Vs Conventional Computers: A Study

on the Larger Scale, Sharjah Center for Astronomy and Space Sciences, University
of Sharjah. Available from
https://www.researchgate.net/publication/323993559_Quantum_Computers_Vs
_Conventional_Computers_A_Study_on_the_Larger_Scale [Accessed 21 October
2022].
7. Goyal,Y. and Sharma, A.(2019).A Semantic Approach for Cyber Threat Predictions
Using Machine Learning. 2019 3rd International Conference on Computing
Methodologies and Communication (ICCMC),2019.Erode,India. 27-29 March
2019.IEEE, 435-438.
8.

13
14