Learning from labelled examples.
WHAT IT REQUIRES
Examples, each with input features and a known correct output.
WHAT THE TWO TASK TYPES ARE
Classification: predicting a category
Regression: predicting a number
WHAT A FEATURE IS
An input value the model uses.
WHAT A LABEL IS
The known answer for a training example.
WHAT THE LEARNING PROCESS IS
Predict, measure error, adjust parameters, repeat.
WHAT A LOSS FUNCTION IS
The measure of how wrong a prediction is.
WHY IT MATTERS ENORMOUSLY
It defines what the model optimises, and therefore what it becomes good at.
WHAT CHOOSING IT WRONGLY CAUSES
A model optimising something other than what you care about.
WHAT GENERALISATION IS
Performing well on data not seen during training.
WHY THAT IS THE ONLY THING THAT MATTERS
Performance on training data is trivially achievable and worthless.
WHAT OVERFITTING IS
Learning the training data, including its noise, rather than the underlying pattern.
WHAT UNDERFITTING IS
The model being too simple to capture the pattern.
WHAT THE BALANCE IS CALLED
The bias-variance trade-off.
