Geoscience Reference
In-Depth Information
CHAPTER
1
Introduction to Statistical
Machine Learning
We start with a gentle introduction to statistical machine learning. Readers familiar with machine
learning may wish to skip directly to Section 2, where we introduce semi-supervised learning.
Example 1.1.
You arrive at an extrasolar planet and are welcomed by its resident little green men.
You observe the weight and height of 100 little green men around you, and plot the measurements
in Figure 1.1. What can you learn from this data?
70
65
60
55
50
45
40
80
90
100
110
weight (lbs.)
Figure 1.1:
The weight and height of 100 little green men from the extrasolar planet. Each green dot is
an instance, represented by two features: weight and height.
This is a typical example of a machine learning scenario (except the little green men part). We
can perform several tasks using this data: group the little green men into subcommunities based on
weight and/or height, identify individuals with extreme (possibly erroneous) weight or height values,
try to predict one measurement based on the other, etc. Before exploring such machine learning tasks,
let us begin with some definitions.
