Lesson video

Hello, my name is Mrs. Holborow, and welcome to Computing.

I'm so pleased you've decided to join me for the lesson today.

In today's lesson, you'll be exploring classification.

What is a class in an AI system, and what common tasks is classification used for? Welcome to today's lesson from the unit Data Science: AI and machine learning.

This lesson is called classification, and by the end of today's lesson, you'll be able to explain how AI systems classify data into categories and how classification underpins many AI tasks.

Shall we make a start? We will be exploring these key words in today's lesson.

Let's take a look at them now.

Class.

Class: a predefined group used to organise data that machine learning or ML developers use to train classification models.

Label.

Label: each piece of data is annotated with one or more labels, which provide information about that data.

Look out for these keywords throughout today's lesson.

Today's lesson is broken down into two parts.

We'll start by explaining what a class is in AI systems, and then we'll move on to describe common classification tasks.

Let's make a start by explaining what a class is in AI systems. AI systems often need to sort data into groups.

This is called classification.

Classification is a type of machine learning.

The groups used to sort data in classification are called classes.

Each category is called a separate class.

Classes help AI systems sort data into useful groups.

Without classes, data would be just a list of numbers or words.

Clear classes make AI systems more useful and accurate.

A simple classification task may involve classifying data into one of just two classes.

For example, predicting whether an email is spam or not.

More complex classification tasks will involve classifying data into one of many categories.

Okay, time to check your understanding.

I have a question for you.

What is a benefit of using classes in AI classification? Is it A, classes reduce the amount of data the AI system needs to process, B, classes make the AI system more unpredictable, or C, classes help the AI system give specific answers, like identifying spam or not spam? Pause the video whilst you have a think.

Did you select C? Well done.

Classes help the AI system give specific answers.

In AI, a class is a category that data belongs to.

For example, photos may be labelled "dog" or "cat," as this example shows on screen.

Emails may be labelled as "spam" or "not spam," as we've seen in previous examples.

Each piece of data is assigned to one class.

Classification uses supervised learning.

The AI model is trained on label examples with known classes.

It is trained to identify patterns that link data to the correct class.

It can then classify new, unprocessed data.

Time to check your understanding.

Classification models must be trained with example data that already has labels assigned by a human.

What labels would you apply to these images? Pause the video whilst you have a think.

Did you manage to identify some labels? Well done.

Let's have a look at a sample answer.

So the most appropriate thing here would be the name of the fruit.

So we've labelled the picture of the orange an orange and then banana and then apple 'cause they just make sense.

The more training data you use, the more accurate the model will be in assigning an item to a correct class.

So you can see here now we've got our classification model, which has the three different types of fruit, the three classes, orange, banana, and apple, and more images have been added to each of those classes to train the model.

Labels guide the AI model during training.

Labels tell the AI model what the correct class should be.

The AI model makes adjustments until the answers match the correct labels.

Poor or biassed labels lead to weak AI performance.

Okay, we're moving on to our first task of today's lesson, Task A.

In your own words, explain what a class is in AI systems. Pause the video whilst you have a go at the activity.

How did you get on? Did you manage to explain what a class is in AI systems? Well done.

Let's have a look at a sample answer together.

In AI, a class is a category or a group that an AI system uses to sort and identify things.

It's like having a set of different folders, and each folder is a class.

For example, in a programme that identifies animals, dog, cat, and bird would all be separate classes.

The AI system uses its training data to decide which class a new image or piece of data belongs to.

Remember, if you'd like to pause your video here and add any detail to your answer, you can do that now.

You can also go back and revisit any previous slides if you need to.

Okay, so we've explained what a class is in AI systems. Let's now move on to describe common classification tasks.

Classification is used every day.

For example, online posts like social media posts or reviews, spam detection, medical checks or diagnosis, or fraud checks used in banking.

Time to check your understanding.

State which description matches each term.

So we've got the terms label and class and the descriptions are: a category that the data can be assigned to and applied to a single piece of data to indicate which class it belongs to.

Pause the video whilst you have a think.

Did you spot it? Well done.

A class is a category that the data can be assigned to and a label is applied to a single piece of data to indicate which class it belongs to.

AI systems help classify online posts.

For example, posts are labelled safe or harmful during training.

The system analyses words, emojis, and images.

AI systems can be used to block harmful content quickly, but they can sometimes remove posts by mistake that are actually safe and okay.

Spam detection is a common use of AI classification.

For example, the AI model is trained on labelled emails marked spam or not spam.

The AI model analyses features such as the subject line, keywords, and sender.

This helps block junk mail, but sometimes, useful emails can get caught by mistake, and they might appear in your junk folder when they're actually valid emails.

AI models are increasingly used in medicine.

For example, AI models can be trained on labelled scans where doctors have marked tumour or healthy.

Patterns are identified in the data and can be used to classify new scans.

This can help detect diseases earlier or confirm a diagnosis, though it must always be double-checked by medical professionals.

AI models are used by banks to spot fraudulent payments.

For example, the AI model is trained on labelled transactions marked as fraud or not fraud.

Features such as the time of the transaction, the location, and the amount spent are analysed.

This helps protect customers, though sometimes, genuine payments can be flagged and blocked.

Classification benefits society and individuals in lots of different ways.

Let's have a look at some of them.

Saving time: repetitive, time-consuming tasks can be automated.

Improving accuracy: AI models can be trained to have high levels of accuracy, which a human may not constantly achieve.

Solving real-world problems such as medical diagnosis and climate change.

Classification can potentially have risks to society and individuals, though.

Let's take a look at a couple.

Outputs may be wrong if the training data is poor quality or not enough training data is provided.

Bias in training data can cause unfair or unrepresentative results.

To improve classification accuracy, AI developers can use large amounts of high-quality training data.

Use a similar amount of training data for each class so that one class is not favoured over another.

Remove any biassed data when training models and try to use training data that represents all of society.

Okay, let's check your understanding.

How can this class training data be improved? Is it A, to add more red apples, B, to add more green apples, or C, to add large apples? Pause the video whilst you have a think.

Did you spot it? Well done.

The correct answer is B.

The current training data doesn't contain a lot of green apples, so if we are going to improve this class, we're going to need to add more green apples to the data.

Okay, we're moving on to our second task of today's lesson.

For Part 1, explain how an AI system could use classification to help an animal researcher sort through images from a wildlife camera in a forest.

Pause the video whilst you have a go at the question.

How did you get on? Did you manage to answer the question? Well done.

You were asked to explain how an AI system could use classification to help an animal researcher sort through images from a wildlife camera in a forest.

An AI system could automatically identify the animals in each image.

The AI system would be trained to recognise different animals and then sort each image into a specific category or class.

The classes could be, for example, deer, bear, fox, and rabbit.

The AI system would process every image and put it into the correct class, saving the researcher a huge amount of time.

For Part 2, what would the animal researcher need to consider when collecting training data for the classification system? Pause the video whilst you answer the question.

Did you manage to think about what training data the animal researcher would need? Well done.

Let's have a look at a sample answer together.

The classification system would need a large amount of training data in the form of animals from existing photographs.

A class would need to be created for each animal and include a range of photographs of this type of animal.

For example, photographs taken from different distances in different locations and at different times of the day.

For Part 3, in your own words, describe one other common classification task.

Pause the video whilst you have a go at the task.

Did you manage to think of a common classification task? Well done.

Let's have a look at a sample answer together.

Remember, you may have described a different common classification task than the one in this sample answer, but that's absolutely fine.

A common example is classifying different species of plants using photographs.

The AI model would be trained on a data set of plants.

Each plant has different features such as the shape of its leaves, colour of flowers, height of the stem, number of petals, et cetera.

The AI model can be trained to identify patterns in these features and then predict which species or type of plant is shown in an image.

Remember, if you want to pause the video here and add any detail to your answers for Task B, you can do that now.

Okay, we've come to the end of today's lesson classification, and you've done a fantastic job, so well done.

Let's summarise what we have learned together in this lesson.

Classification is a type of machine learning task where an AI system assigns data into predefined categories or classes.

AI classification systems are trained on labelled data.

Classification is one of the most fundamental tasks in AI and underpins many real-world systems such as facial recognition.

I hope you've enjoyed today's lesson, and I hope you'll join me again soon.

Bye.