Lecture 5 - Machine Learning

Question 1

What is machine learning?

Answer 1

Making decisions or predictions from patterns observed in data gives “computers the ability to
learn without being explicitly programmed.”
[Arthur Samuels, 1959]

Question 2

People will often talk about the generated code as…

Answer 2

being a “model” a “classifier” or a “regressor”.

Question 3

Machine learning is…

Answer 3

Machine learning is applying an algorithm to data to produce a model that makes predictions.

Question 4

What is supervised learning?

Answer 4

where we know the answer we want for a large number of samples and we want to learn to predict the answer for new ones

Question 5

What is unsupervised learning?

Answer 5

Where we don’t have any particular answers in mind (much more useful for exploratory analysis)

Question 6

What is the attribute we want to learn or predict?

Answer 6

This is called the target variable or often just y.

Question 7

What are the other variables called in a machine learning algorithm.

Answer 7

All the other columns are often called X.

Question 8

Data with an included target are called ?

Answer 8

“labelled data” (sometimes also “ground truth”)

Question 9

Data without an included target are called?

Answer 9

“unlabelled” and are mostly useless for supervised learning

Question 10

In supervised learning we’ll use the labelled training data to?

Answer 10

learn from

Question 11

We define a “cost function” which

helps us define how

Answer 11

“wrong” an answer might be in order to decide how much to “correct” later predictions

Question 12

Examples of supervised machine learning models

Answer 12

Given an email as input, classify it as spam or not-spam
Given data about a house, predict its sale price
Given a date, predict that day’s rainfall
Given a website visitor, predict if they’re likely to sign up
Given a network traffic stream, predict if it’s normal or malicious

Question 13

Two major types of problems we solve with ML?

Answer 13

Classification - Given two (or more) classes, which class does each sample belong to
Regression - Given an input, predict a continuous output variable (e.g. temperature, house price)

Question 14

A “decision boundary”…

Answer 14

Threshold value or tipping point above which we will classify values into class 1 and below which we classify values into class 2

Question 15

A line as a decision boundary in

Answer 15

2-dimensions

Question 16

A surface as the decision boundary in

Answer 16

3-dimensions

Question 17

Dimension n-1 as the decision boundary in n-dimenions

Answer 17

A hyperplane