Studyportal

Question 1. [2 marks]

Describe under which conditions overfitting can happen and provide three possible solutions that can reduce the problem of overfitting. Write no more than 100 words in total.

 

Question 2. [1 mark]

Suppose the features in your training set have very different scales. Explain what alogrithms might suffer from this and what you can do about it. Write no more than 50 words in total.

Question 3. [1 mark]

Explain when logistic regression should be used and why logistic regression is not called “logistic classification”. Write no more than 50 words in total.

Question 4. [1.5 marks]

The generalisation error of a model in machine learning can be expressed as the sum of three very different errors. Name these errors and briefly describe each of them. Write no more than 75 words in total.

Question 5. [1 mark]

Describe the fundamental idea behind Support Vector Machines (SVMs) and explain why it is important to scale the inputs when using SVMs. Write no more than 50 words in total.

Question 6. [2 marks]

The Gini impurity measure is one of the methods used in decision tree algorithms to decide the optimal split from a root node, and subsequent splits. Illustrate how Gini impurity is calculated by providing an example.

Question 7. [1 mark]

Name the main motivations for reducing the dimensonaltity of a dataset and explain what the main drawbacks are. Write no more than 50 words in total.

 

Question 8. [1 mark]

Explain how the TF-IDF measure is calculated and name advantages and disadvantages of this measure. Write no more than 50 words in total.

Question 9. [2.5 marks]

Text processing is one of the most common tasks in many ML applications. Name the most important steps that are involved in data preprocessing and briefly describe each of these tasks. Write no more than 125 words in total.

Question 10. [2 marks]

In order for humans to trust machine learning methods, we need explainability – models that are able to summarize the reasons for the behavior of a machine, gain the trust of users, and produce insights about their decisions. Describe which machine learning algorithm that you
came across comes closest to these requirements and explain why. Write no more than 100 words in total.