Lecture 1: Introduction

to Machine Learning

PUSL3123 AI and Machine Learning

Neamah Al-Naffakh

School of Engineering, Computing and Mathematics

[email protected]
 Today’s Topics
Introduction to Machine Learning (ML)
Lesson learning outcomes: By the end of today’s lesson, you would be able to:

Understand the basic concepts in unsupervised, supervised and

reinforcement learning

Understand how does machine learning work?

 What is Machine Learning?

• What is “Learning”?
• Dictionary definition: “the process or experience of gaining knowledge or skill
through study, experience, or being taught”.

• An example of learning is a student understanding and remembering what they've

been taught, learning how to drive, learning how to recognise digits

• Can a Machine “Learn”? E.g., learn from data or learn from past experience?
 What is Machine Learning?
• (Arthur Samuel, 1959) concluded that “a computer can be programmed so that it will
learn to play a better game of checkers than can be played by the person who wrote
the program.”
With algorithms
❑ Learn from Experience E
❑ Achieve some task/s T
❑ Improve their performance P

Example: SPAM email filtering

• Collect emails which have been labelled as spam or non-spam –E
• Classifying emails as spam or non-spam and putting them into the right folder –T
• Calculate the percentage of emails correctly classified as spam or non-spam -P

Machine Learning is the study of computer algorithms that improve automatically

through experience.
Why is Machine Learning Important?
Applications of Machine learning
 Applications of Machine learning
✓Image Recognition
identify objects, persons, places, digital images, etc

✓Speech Recognition
Speech recognition is a process of converting voice instructions into text, and it is also known as "Speech to text",
or "Computer speech recognition.“. For example, Search by voice in Google

✓Traffic prediction:
Using Google Maps to find out the correct path with shortest route and predicts the traffic conditions. This can
be achieved with the help of two inputs:
o Live GPS information of the vehicle
o Average time has taken on past days at the same time.
 Applications of Machine learning

✓ Medical Diagnosis

Machine learning is used for Disease identification. For example, It helps in finding brain
tumors and other brain-related diseases easily.

✓ Self-driving cars

One of the most exciting applications of machine learning is self-driving cars. For example,
Tesla (i.e., car manufacturing company) is working on self-driving car by using unsupervised
learning method to train the car models to detect people and objects while driving.
Types of Machine Learning
 Supervised Machine Learning

• Supervised Learning: given samples of input data and output data, machine to learn relationships

between them, and then be able to predict output from unseen input data

• The machine learns under the guidance of labeled data i.e. known data

• It is based upon the training dataset, and it improves through iterations.

• Classifying massive data with supervised learning is difficult

Supervised Machine Learning
 Supervised Machine Learning
For example, the images of fruits can be tagged as apples, bananas, grapes or Oranges.
Supervised Machine Learning
Unsupervised Machine Learning
• Unsupervised Learning: Only given input data without known output data, machine to
learn latent/hidden relationships within the data and make decisions.

• It’s a self-organized learning algorithm (no labelled data is required).

• Algorithm would be fed a lot of data and given the tools to understand the properties of the data.

• Unsupervised Learning an interesting area is that an overwhelming majority of data in this world is
unlabeled
 Reinforcement Learning
• Reinforcement Learning: get machine to learn by itself via trial and error, interact with
environment, and take actions to achieve a final goal.

• Reinforcement learning comprises three significant components: agent, environment, and

actions.

▪ The Agent: the learner/decision-maker

▪ The Environment: anything the agent interacts with

▪ The Actions: what the agent does

Reinforcement Learning
How Does Machine Learning Work?

• Choose and Prepare a Training Data Set

• Select an Algorithm to Apply to the Training Data Set

• Train the Algorithm to Build the Model

• Use and Improve the Model

 1- Choose and Prepare a Training Data Set

• Training data is also known as training dataset, learning set, and training set

• Training data is the initial data used to train machine learning models.

• Without high-quality training data, even the most efficient machine learning algorithms will

fail to perform.

• Training data can be classified into two categories: labeled data and unlabeled data.
 Labelled data VS Unlabelled data?
• Labelled data

• It is a group of data samples tagged with one or more meaningful labels. Labels used to
identify specific characteristics, properties, classifications, or contained objects.

• Labelled training data is used in supervised learning.

• Data labelling is a time-consuming process as humans need to tag or label the data points.

• Unlabelled data

• It is raw data or data that's not tagged with any labels for identifying classifications,
characteristics, or properties

• Unlabelled data used in unsupervised machine learning

 2- Algorithm Selection
• The selection process of machine learning algorithm depend on a few aspects

• Whether the use case is prediction or clustering

• How much data is in the training set

• The nature of the problem the model seeks to solve

3- Train the Algorithm to Build the Model

• Setting model variables and parameters to more accurately predict the appropriate
results.

• Using a variety of optimization methods depending upon the chosen model.

 4- Use and Improve the Model
• Feed new data to the model as a means of improving its effectiveness and accuracy
over time.

• Feature Engineering which create new features from the existing ones.

• Feature Selection that helps to identify the most useful features in the dataset.

• Try Multiple Algorithms to identify which ones work best for data and then use that
information to improve the accuracy of models.

• Adjusting Hyperparameters such as number of layers in a deep neural network).

• If the prediction and results don’t match, re-train the algorithm multiple times until
getting the desired outcome
Conclusion
Reading List
• Christopher M. Bishop: Pattern Recognition and Machine Learning,Springer, 2006.

• Kevin Murphy, Machine Learning: A Probabilistic Perspective, MIT Press, 2012. (e-book available via Primo)

• Richard S. Sutton, and Andrew G. Barto, Reinforcementlearning: An introduction. 2ndEdition. MIT press, 2018.
(e-book available via Primo)

• Tom Mitchell, Machine Learning, McGraw-Hill, 1997.

• Ian Goodfellow, YoshuaBengioand Aaron Courville, Deep Learning, MIT Press, 2016. (online version available).