ML Overview

Question 1

What are the 6 stages of the ML Pipeline

Answer 1

1. Define the problem
2. Data collection
3. Data preprocessing
4. Data modelling / machine learning
5. Model evaluation
6. Model application (on new/unseen data)

Question 2

Define ML

Answer 2

Input data and answers and output rules that can be applied by the computer to new situations.

Question 3

What are the 4 types of data sets?

Answer 3

1. Record
2. Graph and network
3. Ordered
4. Spatial, image, multimedia

Question 4

What are data objects?

Answer 4

1. Make up datasets
2. Represent the entity being measured
3. Are a row in a database
4. Aka entities, instances, points, samples, tuples, patterns, vectors, examples

Question 5

What are data attributes?

Answer 5

1. Describe the data objects
2. Are the columns in the database
3. Aka features, variables, dimensions, predictors

Question 6

What are the five types of data attributes

Answer 6

1. Nominal - categories
2. Binary - 0,1
3. Ordinal - meaningful order but magnitude between successive values is not necessarily meaningful
4. Interval scaled - equal sized units, ordered, no tire zero
5. Ratio scaled - as per interval but with a true zero

Question 7

3 major categories of machine learning

Answer 7

Supervised
Unsupervised
Reinforcement

Question 8

Define supervised ML

Answer 8

Learning with a labelled training dataset

Question 9

Define unsupervised ML

Answer 9

Learning patterns in unlabelled data

Question 10

Define reinforcement ML

Answer 10

Learning based on feedback and reward

Question 11

2 types of supervised ML

Answer 11

Classification
Regression

Question 12

2 types of unsupervised ML

Answer 12

Clustering
Anomaly detection

Question 13

2 types of reinforcement learning

Answer 13

Game play
Control

Question 14

3 regression methods

Answer 14

Linear regression, polynomial regression
Ridge regression, LASSO, elastic net
Artificial neural networks

Question 15

4 classification methods

Answer 15

Logistic regression
K nearest neighbours
Support vector machines
Decision trees

Question 16

3 things that determine which ML approach to take

Answer 16

Data
Computing requirements
Interpretatability

Question 17

What 3 aspects of the data determine which ML approach to take

Answer 17

1. Size of data (more model parameters need bigger datasets)
2. Feature types
3. Linear or non-linear relationships

Question 18

What 2 computing requirements determine which ML approach to take?

Answer 18

1. Training period available
2. Production performance requirements