2-mlconcepts

Question 1

What is an instance?

Answer 1

An instance is an individual, independent example of a concept. (Rows)

Question 2

What is an instance also known as?

Answer 2

Instances are also known as exemplars

Question 3

What is an attribute?

Answer 3

Attributes are the measuring aspects of an instance. (Columns)

Question 4

What are attributes also known as?

Answer 4

Attributes are also known as feature

Question 5

What are concepts?

Answer 5

Concepts are the things we aim to learn, generally in the form of labels or classes

Question 6

What is a supervised learning method?

Answer 6

Supervised learning methods have prior knowledge of a closed set of classes
and set out to discover and categorise new instances according to those
classes

Question 7

What is an unsupervised learning method?

Answer 7

Unsupervised learning methods do not have access to an inventory of classes, and instead discover groups of ‘similar’ examples in a given dataset

Question 8

What is “strong” unsupervised learning?

Answer 8

Strong unsupervised learning dynamically discovers the “classes” (implicitly derived from grouping of instances) in the process of categorising the instances

Question 9

What is “weak” unsupervised learning?

Answer 9

Weak unsupervised learning categorises instances as certain labels without the aid of pre-classified data

Question 10

What are nominal quantities?

Answer 10

Nominal quantities are distinct symbols, serving only as labels or names. No relationship is implied on nominal values

Question 11

What are ordinal quantities?

Answer 11

A set of labels with an explicit total order.

Question 12

What are numeric quantities?

Answer 12

Numeric quantities are real valued attributes

Question 13

How can nominal values be converted to numeric values?

Answer 13

1. Map categories to numbers (creates artificial ordering)
2. One hot encoding (increases dimensionality)

Question 14

What is feature standardisation?

Answer 14

Feature standardisation rescales features to be distributed around a 0
mean with a unit standard deviation

Question 15

What is feature scaling?

Answer 15

Feature scaling rescales features to a given range. For example, min-max scaling rescales values between 0 and 1 using the minimum and maximum feature value observed in the data

Question 16

What is discretization?

Answer 16

Discretization groups numeric values into a pre-defined set of distinct categories. This can be done through binning the dataset.

Question 17

What is equal width discretization?

Answer 17

Equal width discretization, partitions the data set into n bins of width (max - min)/n

Question 18

What is equal frequency discretization?

Answer 18

Equal frequency discretization, partitions the data set into n bins, each with an equal number of values. Requires dataset to be sorted before binning.

Question 19

What ways can numeric values be discretised?

Answer 19

1. Equal width discretization
2. Equal frequency discretization
3. Clustering (K-means)

Question 20

What is linear regression?

Answer 20

Linear regression is a model, that estimates a real value (numeric class) through a linear combination of real valued attributes and attribute specific weights

Question 21

What is the ordinary least squares learning method?

Answer 21

The ordinary least squares learning method, minimises the sum of squared errors in linear regression

Question 22

What is the sum of squared errors (SSE)?

Answer 22

The sum of squared errors is the sum of the square of the difference between the actual and predicted values