Report Writing

[GiteonCaulfied]

Hi @amartinhuertas @rhyshawkins @sghelichkhani ,

I've just added the report file Report.pdf to the repo. The structure of report follows the LaTeX project report templates provided in the reference with a few changes, including removing the Appendix and Evaluation chapter and separate the "Main Content" chapter into two different chapters for 1D Geoid problem and 2D mantle convection problem. (The evaluation chapter is removed because I plan to document and evaluate the result in these two chapters separately)

Feel free to make suggestions if you think the current structure has some problems or more things could be added to the report.

[amartinhuertas]

Hi @GiteonCaulfied,

by now, a couple of comments:

1. Can you also push to the repo the latex sources of the report? I think it would be good to track them down as well here.
2. Can you add some references to the current report using BibHandler.jl + bib file so that you can familiarize yourself with reference handling in Latex?

It would be also good if we can get the background from @sghelichkhani and @rhyshawkins so that you can start populating some sections of the report.

[GiteonCaulfied]

Hi @amartinhuertas ,

I've added the latex sources to the repo, which is a folder called Report LaTeX. Also, I generated the references.bib file using the BibHandler you provided and add some references to the "Related Work" chapter of the report without any context. It works smoothly and now I can see how powerful this BibHandler is. Thanks for sharing it!

[GiteonCaulfied]

Hi @amartinhuertas,

I reconstruct the structure of the report a little bit by removing the "Related Work" chapter and put it as a subsection in the "Background" chapter. In this case, the logic flow of the report will be like:

• Introduction

  • General background for Geoid problem and Mantle Convection
  • General Introduction (e.g. I use neural network to reduce the computation complexity of these two problems by...)

• Background

  • What is Neural Network? What is FNN, CNN and LSTM?
  • Related work to solve these two problems using Neural Network

• Solving Geoid problem using FNN

• Solving mantle convection using ConvAE along with FNN or LSTM

• Conclusion and Future Work

Also, I write a few paragraphs in the Geoid chapter. For the Mantle Convection chapter, I am thinking about whether my current result for the 2D problem can be put into the report since this data set uses adaptive timestamps and Sia is currently generating a new one.

The updated report and latex sources have been added to the repo.

[GiteonCaulfied]

Hi @amartinhuertas,

I've added more words to the report, including:

• A short paragraph in the Introduction (Chapter 1) explains the structure for the rest of the report
• Background for the NN techniques used in this study
• Complete chapter for solving mantle convection problem (except for LSTM since the current result may have underfitting problem and I need to test with Sia's new data set)

The updated report and latex sources have been added to the repo. I will try to add some paragraphs in the "Related works" section of Chapter 2 and write some background in Chapter 1 later.

UPDATE: I've added a few paragraphs to the "Related works" section of Chapter 2. For now I will leave the report for a few days since Sia has generated new data.

[amartinhuertas]

Hi @GiteonCaulfied ... for my knowledge. Which is the deadline for a first complete draft of the report?

[GiteonCaulfied]

Hi @amartinhuertas,

The deadline is Week 8 Friday 29 September, 16:30. And it should be only submitted to the supervisor.

[amartinhuertas]

I was actually meaning the deadline for submitting the report to the committee that is going to evaluate it.

[amartinhuertas]

Ok, I now see that in Wattle we have:

Week 12 Friday 27 October, 16:30 | Written report submission | 45% to 60%

All good.

[amartinhuertas]

I would thus wrap up with these final set of experiments, and focus in the report. We have one month from now on, but one month flies ...

[GiteonCaulfied]

Hi,

I've restructured the mantle convection simulation chapter of the report by splitting this chapter into three sections:

• Mantle Convection Simulation on Limited dataset
• Mantle Convection Simulation on Larger dataset
• Mantle Convection Simulation on Interpolated dataset

Each of these section includes a short introduction to the dataset, result on ConvAE, result on FNN and result on LSTM as the subsections. Figures for these subsections are added as well.

[amartinhuertas]

As a general comment, I see some of the figures in the report are missing an x-axis label, y-axis label, and title. We should have these in all figures.

[GiteonCaulfied]

Hi @amartinhuertas ,

I've added some of the missing x-axis labels and y-axis labels. For the temperature fields, I only added the x-axis label and y-axis label for an example figure and since adding labels to all of them can make the figures looks a bit messy (this is explained in the report as well for clarity). As for the other figures without labels in the Geoid chapter, I haven't added those since I am not sure what to label with due to the lack of background.

The PCA in the report is now renamed to its correct name POD(Proper orthogonal decomposition). Previously I thought it could be called as PCA since they are highly similar and they all use eigenvalues for analysis. But for the sake of clarity I think it's better be called as its original name POD instead.

[GiteonCaulfied]

Hi @amartinhuertas @rhyshawkins,

An updated version of the report has been pushed to the repo.

I made a few changes to different chapters according to the comment I received yesterday. I also added some Geoid visualization figures to the Geoid Chapter using the code from yesterday's pull request. The figure looks great as expected and when I was writing this comment I just noticed that they are missing x-axis labels and y-axis labels, I will add them back in my next commit.

Some figures explaining the workflow of ConvAE, FNN and LSTM have been added to the poster but not the report and I will do this later.

UPDATE: these figures have been added to the report, along with the missing x-axis labels and y-axis labels