Machine Learning: How Computers Learn, and Tools to Try It

Topic: Y2 Machine Learning simulation tools

What machine learning is and how to try it yourself, explained step by step: everyday examples, supervised, unsupervised and reinforcement learning, preparing data, ML libraries, building a model with Teachable Machine and Machine Learning for Kids, and the ethics of bias, fairness, privacy and accountability; how ML fits into AI.

View resource Free to join

What's inside

When Jumia suggests something you might like, when your email moves a scam into spam, or when your phone's keyboard guesses your next word, nobody wrote a rule for each of those cases. The computer worked it out from lots of examples. That is machine learning, and with free online tools you can train a simple model yourself in a few minutes.

SUMMARIZED CONTENT

Year 1: No content on machine learning in Year 1.
Year 2: What machine learning is, its three types, preparing data, ML libraries, building a model with simulation tools, and ethical considerations.
Year 3: Machine learning as part of artificial intelligence (see the topic on AI).

Year 2: Machine learning and how to try it

What machine learning is:

Think about how you improve at maths: the more problems you practise, the better you get. Machine Learning (ML) is the part of Artificial Intelligence that teaches computers to learn and improve from data, without being told exactly what to do for every case. The more good examples a model sees, the better its predictions become.

You meet ML every day: social media feeds learn what posts you like; online shops like Jumia recommend products from your searches; voice assistants such as Siri or Google Assistant get better at understanding your voice; spam filters block unwanted email; and weather forecasts use ML to analyse weather data.

The three types of machine learning

Supervised learning is like studying with an answer key. The computer learns from labelled data, where the correct answers are given, for example pictures labelled "apple" or "orange". After enough examples, it can label new pictures itself. It is used for image classification (identifying animals) and predicting house prices from size and location.

Unsupervised learning is like being handed a pile of mixed-up items and asked to sort them into groups, with no labels. The computer finds patterns on its own. A common task is clustering: grouping data points that are similar (close together) and separating those that are different. Uses include grouping customers with similar shopping habits, finding objects in images, organising documents, and spotting unusual data that might be fraud (anomaly detection). Common methods are K-means clustering (into a set number of groups) and hierarchical clustering (merging similar items step by step).

Reinforcement learning is learning by trial and error, like training a dog: a treat for catching the stick, nothing for missing it. The agent (the learner) acts in an environment, takes actions, makes observations about its situation, and receives rewards. A robot learning a maze gets points for moving closer to the exit. It is used to train self-driving cars, robots and drones.

Open the full resource free with a TudHub account.

More on Machine Learning simulation tools

Other resources you can find on TudHub.