4 Understand machine learning
Machine learning involves feeding large amounts of data into an algorithm, which then processes this data to find patterns and relationships.
For example, if you provide an AI model with data about your past vacuuming habits and the layout of your home, the model that results could be used to program a robot vacuum to sweep your carpet in a similar pattern to you. Now imagine many of these robots were sold, transmitting all their vacuuming data back to the developer. Using machine learning to detect patterns could discover even more efficient movement patterns and optimise paths for all robot vacuums with the same design.
In the first video, you’ll gain a better understanding of what machine learning is.
Download this video clip.Video player: Vidoe 6
Transcript: Vidoe 6
Voice Over:
You might have heard the terms’‘machine learning’ and ‘AI’ used interchangeably, and you may be wondering, ‘What is the difference?’ Well, there is a difference, and it is important to understand how they are different.
Artificial Intelligence refers to the intelligence exhibited by software and machines. Machine learning, on the other hand, is a subset of AI.
This means that while all machine learning is AI, not all AI involves machine learning.
Machine learning is a type of AI where a machine learns from the data it has been given and can identify patterns in that data.
It is also the process by which machines learn from data and improve their performance over time.
It uses different types of techniques such as supervised learning, unsupervised learning, and reinforcement learning.
The term ‘Machine Learning’ was introduced by Arthur Samuel in 1959. Samuel, an American pioneer in the field of computer gaming and AI, created a program that played draughts against itself.
The machine analysed the game through repetitive plays, identifying strategies to win and avoid losses. Through playing the game and learning what to do to win, it began to detect recurring patterns and would increasingly follow a pattern.
To achieve this, Samuel had to use mathematics. Concepts such as linear algebra, calculus, probability, and statistics play a crucial role in understanding how machine learning algorithms learn from data and make predictions.
These mathematical concepts help in optimising the performance of the model, understanding the relationships within the data, and making accurate predictions.
This draughts game is a prime example of a machine learning program. It was learning from its experiences, replicating successful results in other games, and refining its performance over time.
Need further clarity? Let’s consider the analogy of learning to ride a bicycle.
When you first learn to ride a bicycle, you might have started with training wheels. These training wheels are like the initial data set that we feed into the machine learning model. They provide the basic guidance and stability that the model needs to start learning. As you practise more and more, you start to understand how to balance, when to pedal, and how to steer. This is like how a machine learning model begins to recognise patterns and relationships in the data during the training process.
Eventually, the training wheels come off. Now, you are not just riding the same bike in the same way. You are adapting to different situations - maybe you are riding on a hilly path or navigating through a crowded park.
Similarly, a machine learning model uses the patterns it has learned to adapt to new data and make accurate predictions or perform tasks it was designed for.
Just like how you can ride different bicycles after you have learned the skill, a machine learning model can apply its learning to different but related problems. The bicycle you ride today might not be the one you learned on, but the skill you learned transfers across.
The fundamental concept of machine learning is when a model learns from data, identifies patterns, and uses these patterns to make predictions or decisions.
Just like Samuel’s draughts program, machine learning models improve their performance over time through continuous learning and adaptation.
So, remember, all machine learning is AI, but not all AI is machine learning, and machine learning is broken into different types such as supervised, unsupervised, and reinforcement learning.
Vidoe 6
Interactive feature not available in single page view (see it in standard view).
Machine learning is an AI technique that uses algorithms to create predictive models. These models are validated against known data, measured by specific metrics, and adjusted as needed. This process of learning and validation is called training. Through retraining, machine learning models improve overtime. There are also key methodologies on how to train models, for example, supervised learning, unsupervised learning, and reinforcement learning.
In the next video, you’ll learn the difference between these methodologies, and be introduced to their combination in layers, called deep learning.
Download this video clip.Video player: Video 7
Transcript: Video 7
Ever wondered how you learned to ride a bike or play an instrument?
It is all about learning from experience, correct? Well, machines can do that, too! They can learn in three main ways: supervised learning, unsupervised learning and reinforcement learning.
Let’s break it down each with a simple analogy: learning to play football.
In the first scenario, you have a trainer who supervises you and teaches you all the rules.
They carefully explain that when the ball enters the goal, you earn a point, and when the ball goes out of play over the sideline, it must be thrown in by hand.
They explain all the rules, how many players there are, how long the game lasts, and so on. In this case, you have a person supervising you and teaching you all the rules of the game.
This is like supervised learning. In the second scenario, you are on your own.
You start attending games every Thursday and Sunday. Initially, you are puzzled.
Why are players using their hands from the sidelines in a game primarily played with feet?
Why is the crowd cheering when the ball hits the net, and why do they groan when it does not? But as you keep watching, you start recognising patterns and understanding the game’s dynamics.
This is like unsupervised learning. Now, let’s consider a third scenario, which is like reinforcement learning.
Imagine you are practising penalty kicks.
Each time you score a goal, you feel a sense of achievement, which is a positive reward. Each time you miss, you feel disappointment, which is a negative reward.
A thumbs down appears when the ball hits the goal post The ball begins going in the net. Four stars light up.
Over time, by trying different ways of kicking the ball and learning from the rewards, you improve your ability to score penalty kicks.
So, remember, all machine learning is AI, but not all AI involves machine learning.
Machine learning can be broken down into these types: supervised, unsupervised, and reinforcement learning. But we can go even deeper.
There are advanced forms of machine learning that mimic the human brain’s own method of learning.
They take the concepts of supervised, unsupervised, and reinforcement learning and apply them on a much larger scale. This is called deep learning.
Just as neurons in the brain are connected to form a vast network, deep learning uses artificial neural networks with several layers - hence the term ‘deep’.
These networks can learn and make decisions on their own. Intriguing, isn’t it?
Video 7
Interactive feature not available in single page view (see it in standard view).