💡 Topics

In this lesson, we will cover the following:

• Generalization and machine learning

🎓 Intended learning outcomes

At the end of this lesson, students are expected to:

• be able to describe generalization and why it is important to ML
• explain the assumptions we make when training a model
• describe overfitting and underfitting
• understand the differences between goals of ML & optimization

In this module, we will look at machine learning through the perspective of generalization. Why? Because generalization can help tell us whether the ML model we’ve built is actually useful.

Imagine this scenario: the incident with your fridge in the previous module has got you thinking about rotten fruit, and in particular, how to pick a ripe avocado. Using your newfound ML & optimization skills, you set to work building a regression model that predicts the number of days until peak ripeness of an avocado and you turn it into an App.

It was easier than you imagined. You went to the supermarket and bought 50 avocados, took a photo of each, and then checked each every day to see when it was at peak ripeness. The trickiest part was boiling down the visual appearance of the avocado into a single parameter you dubbed squelchiness that nicely relates to the number of days until the avocado is ripe.

You used linear regression to fit a polynomial to the data you collected, and you were pleased to see a nicely fitted model.