Unsupervised Learning and Probabilistic Inference

Question 1

In ML what do we start with

Answer 1

Observations (dataset) of a system or phenomenon

Question 2

What is the aim when given a dataset in ML

Answer 2

To discover patterns which leads to discovering general knowledge that can be used

Question 3

What is a key limitation of learning from data

Answer 3

A model can only learn what is present in the data

Question 4

What else limits a machine learning model beside data

Answer 4

The choice of model

Question 5

What matters more than the quantity of data

Answer 5

The quality of data

Question 6

What questions should we asked about data quality

Answer 6

Is the data representative of the underlying phenomenon

Are the features sensible

Have we included all the features we need to

Do we have observations from all regions of the input space

Question 7

What do rows in the data matrix represent

Answer 7

Observations

Question 8

What do columns in the data matrix represent

Answer 8

Features

Question 9

What is the aim of unsupervised learning

Answer 9

Not to make predictions, but to understand and or gain insight into the phenomenon itself

Question 10

What does unsupervised learning try to find

Answer 10

Internal relationships and patterns

Question 11

How are observations analysed in unsupervised learning

Answer 11

By comparing observations to each other

Question 12

What key question does unsupervised learning ask abut data

Answer 12

Do we have different high level types of observations (clusters)

Question 13

What is another key goal of unsupervised learning besides clustering

Answer 13

Dimensionality reduction

Question 14

What is dimensionality reduction

Answer 14

Representing the data matrix in a reduced form while preserving important info

Question 15

Why do we do dimensionality reduction

Answer 15

To make the dataset more efficient and easier to analyse without losing information

Question 16

What is K-Means clustering used for

Answer 16

To group observations into K clusters based on similarity

Question 17

What are the steps of K-Means clustering

Answer 17

1. Choose the number of clusters (K)
2. Select K initial centroids (starting points)
3. Assign each point to the nearest centroid
4. Update centroid positions based on the assigned points
5. Repeat steps 3 and 4

Question 18

What does K-Means clustering stop

Answer 18

When convergence is reached (centroids stop changing significantly)

Question 18

How do you perform one full iteration of K-Means clustering

Answer 18

Assign each data point to the nearest centroid

recalculate each centroid as the mean of its assigned points

Question 18

What calculation is used when updating a centroid

Answer 18

The mean of all points assigned to that cluster

Question 18

How do you assign a point to a centroid in K-Means

Answer 18

Choose the centroid with the smallest distance to the point

Question 19

What is supervised learning

Answer 19

Comparing observations to distinguish them in terms of a target

Question 20

What is unsupervised learning in contrast

Answer 20

Learning without labels, aiming to get an insight into the dataset

Question 20

What is the aim of supervised learning

Answer 20

To distinguish between inputs such that this helps distinguish between targets

Question 21

What is the key difference between supervised and unsupervised learning?

Answer 21

Supervised learning uses targets; unsupervised learning does not.

Question 22

What is probability used for in ML?

Answer 22

To deal with uncertainty.

Question 23

How is probability interpreted?

Answer 23

As a degree of belief.

Question 24

What defines a probabilistic model?

Answer 24

The joint distribution.

Question 25

What probability rules are important?

Answer 25

Sum rule, product rule, and conditional probability.

Question 26

What is marginalisation?

Answer 26

Propagating belief or uncertainty by summing over variables.