Overview

Supervised learning

In supervised learning we are given pairs \((x_i, y_i)\) and we want to predict \(y\) from \(x\) for new inputs.

Pages

Always look at the raw data	The weather data, and what a pair plot hides
Loss minimization	Function family, loss, optimizer
Likelihood maximization	The same machine, from a distribution
Linear regression	Simple and multiple, on the weather data
Exercises

Goals

The goal of this week is to

1. understand linear regression from two equivalent perspectives: as a machine that tunes its own parameters to minimize the mean squared error loss, or to maximize the log-likelihood,
2. translate the blackboard example of linear regression into code, and understand the concept of a data generator, and
3. know how to perform (multiple) linear regression on a given data set.