Examiner comments

[daffidwilde]

The following is a verbatim list of comments left by my examiners. These comments were drawn from the notes of my internal examiner.

Chapter 1: Introduction

• Restructure Chapter 1 to improve the narrative and the connection between the different Chapters and how everything connects to the title of the thesis.

Chapter 2: Literature review

• Section 2.5: Discuss the connection of the literature to Chapter 3 a bit more clearly. Maybe re-arrange to put things in order.

Chapter 3: EDO

• Figure 3.1: Maybe discuss the limitations of this paradigm. Can we apply it universally? Are there specific conditions that need to be satisfied.
• Section 3.2: It is a long list of parameters. Is there an optimal list of parameters? Enhance a bit the discussion around the choice of parameters. Later in the section (page 66) you seem to work without the optimal list of parameters. This seems to be a big limitation and needs to be discussed at the end of Chapter 3.
• Figure 3.6: What defines “Best” in the Figure. Can you discuss this in the document?
• Section 3.3: You use clustering to show a case study/example of the EDO algorithm. Is the process completely different for different statistical procedures, i.e. classification? What parts are similar and what are completely different? It will be helpful to discuss it a bit, so that someone interested might see how to apply this technique on a different procedure.
• Page 72 (middle of the page): This sounds like you are trying something, and then modifying it to make it work in a random way. You explained it much better during the viva. So maybe add a bit of discussion there, where -f makes sense.

Chapter 4: k-modes initialisation

• Page 82: Type a proper Theorem/Proposition that you are proving.
• Page 83 Theorem 4.1: Make it clear is a well-known Theorem.
• Page 102: The process that yields 35000 datasets. It is good to give the computational time to run this algorithm. It is also good to see if you can run it with different set of parameters or even run a sensitivity analysis to check if deviations in the parameter values affect the performance of the algorithms.

Chapter 5: COPD study

• Generally, on this Chapter you need to write more clearly how and what aspects of EDO gets involved. Also, at the end of the Chapter do a bit of reflection on the advantages of using EDO.
• Page 135 (last sentence of section 5.4.1): You may want to enhance what you write here by explaining what that means for the model and what it means for the system. Is it better? Is it worse?
• Page 140: Can you discuss the results in Tables 5.3, 5.4 in connection with the observations from Figures 5.2 to 5.6 at the beginning of the Chapter.

[daffidwilde]

• Replace the hyphens in "best, most-middling, and worst-performing"