Exam

Question 1

What is a classification problem

Answer 1

A problem that requires machine learning algorithms that learn how to assign a class label to examples from the problem domain

Question 2

What is a regression problem

Answer 2

A problem that learns to predict continuous variables

Question 3

What algorithms are used for Regression Problems?

Answer 3

• Linear Regression
• Support Vector Regression
• Regression Tree

Question 4

Give an example of a Classification problem

Answer 4

Getting a machine to classify different images such as the difference between apple[1,0,0], banana[0,1,0] and cherry[0,0,1]

Question 5

What is Underfitting?

Answer 5

When a model cannot capture underlying trend of the data

Question 6

Why does Underfitting occur?

Answer 6

Algorithm does not fit/ Not enough data

Question 7

What happens with the Bias& Variance in Underfitting

Answer 7

High bias and low variance

Question 8

What is Bias?

Answer 8

Assumptions made by a model to make a function easier to learn

Question 9

What is Variance?

Answer 9

Training data obtains a low error, and then changing training data obtains a high error

Question 10

How to prevent Underfitting

Answer 10

Increase model complexity
Increase number of features (feature engineering)
remove noise
Increase epochs

Question 11

What is overfitting?

Answer 11

Trained with a lot of data, the model starts to learn from the noise and inaccurate data entries. The model has too much freedom and builds an unrealistic model

Question 12

What is overfitting in terms of variance and bias

Answer 12

High variance and low bias

Question 13

How to reduce overfitting

Answer 13

Increase training data
reduce model complexity
early stopping
L1&L2 regularization
Dropouts if neural network

Question 14

What is regularisation

Answer 14

the technique of calibrating machine learning models to minimize the loss and prevent over or underfitting

Question 15

What noise mean?

Answer 15

The data points in a dataset that don’t really represent the true properties of your data

Question 16

What does Bias mean in terms of regularisation?

Answer 16

the difference between the actual and predicted values Less consideration to data pattern = oversimplified and underfit models

Question 17

What does Variance mean in terms of regularisation

Answer 17

Measure of flexibility in the model. Decides how sensitive the model is to change based on the patterns in the input data

Question 18

What happens to the training and testing error when the bias is high

Answer 18

They will also be high

Question 19

What happens to the training and testing error when the variance is high

Answer 19

They will be low

Question 20

Name the two main types of regularization techniques

Answer 20

Ridge and Lasso Regulation

Question 21

What is Ridge regularisation

Answer 21

Modifies over or underfitted models by adding the penalty equivalent to sum of the squares of the magnitude of coefficients

Question 22

what is Lasso Reggression

Answer 22

Modifies the over fitted/underfitted models by adding a penalty = to the sum of the absolute values of coefficients

Question 23

What is Dropout In regularisation

Answer 23

Randomly selected neurons are ignored during training. Dropped out randomly. therefore their contribution is temporally removed

Question 24

What happens as a neural network learns

Answer 24

Weights settle into their context within the network. Weights are tuned for specific features, providing some specialization. Neighboring neurons come to rely on this specializations which can result in a fragile model too specialized for training the data.

Question 25

How does dropout help with overfitting

Answer 25

1. Neurons cannot rely of one input as it may dropout at random - this reduces Bias due to over-relying on one input
2. neurons will not learn redundant details of inputs

Question 26

The concept of concept attainment requires the following 5 categories

Answer 26

1. identify task
2. nature of examples used
3. validation procedure
4. consequences of categorizations
5. nature of imposed restriction

Question 27

what is an decision tree

Answer 27

A supervised learning algorithm (regression and classification) Tree structure with roots, nodes and branches like a flowchart

Question 28

Advantages of decision trees

Answer 28

• easy to interpret
  -no data preparation required
  -more flexible

Question 29

Disadvantages of decision trees

Answer 29

-prone to overfitting
-high variance
-more costly