# Get Started

Please spend about a month studying the tools to be used in your FYP. For example, if you use neural networks, you can find a book in NTU library to read about the basics (you don’t need, but are welcome, to learn ALL about the tools or advanced materials). You could experiment it in Matlab with some tutorials, toolboxes and manuals.

You also need to read some research papers published on your FYP. You need to understand what has been done and how it has been done, so that you know what to do in your project. Later you need to write about prior work in your FYP report. In addition, while reading research papers, pay attention to how the papers are written, so that you know how to carry out your FYP work and write your FYP report properly. Your FYP report should be suitable for submission for publication in a journal (preferred) or conference.

It is not necessary (and can even be bad) to read past year FYP reports.

You should aim at 1 or both of the following goals, if you are interested in an “A” grade.

1. Practical goal

To show a method/algorithm (published by someone in research papers or books, either unchanged or modified by you) work for a specific situation. Your results may be of practical importance, for example, you could really use the results to make money in a particular stock or index. This goal may be hard to achieve (no FYP student has achieved this so far).

First, reproduce the results in a published research paper using Matlab and the same data used in the paper. Second, make some modifications to the algorithms in the paper and try to improve the results for the same data used in the paper. To do these, you should first make sure the data used in the paper can be freely downloaded. Avoid comparing to old papers (e.g., before 2010).

It would be less desirable to simply do something, without comparing with published research paper(s), or simply reproduce someone else’s results without any improvements.

During your computer simulations, whenever appropriate, calculate averages and standard deviations.

Standardize the result presentations. For example, an accuracy of 53.5973% usually does not make sense and 53.6% would be sufficient.

Use tables and figures to present and compare results, rather than bullets. For comparison, each table should have multiple columns/rows and each figure should have more than 1 curve/dots.

Writing Final Report

Do a spelling and grammar check.

Do not copy anyone else’s stuff, including sentences, paragraphs, figures, pictures, pseudo-codes, etc (it is against copyright law to do any of these). Write your own sentences. Draw your own figures. Do not include any scanned or copy-righted materials. Violations will result in heavy penalties.

For each method/algorithm used: if it is invented by you, say “we propose…”. If it is NOT invented by you, write “… was proposed by … (give references. Use past tense). If it is your modification of someone else’s work, say exactly what the modifications are and give references. You must review all existing work directly related to yours. Do not include other people’s work that is not directly related to your results.

Use simple present tense to describe your work. Use simple past tense to describe already published work.

Define EVERY acronym (e.g., TSP) when it first appears (e.g., traveling salesman problem (TSP)). Don’t define any acronyms that you don’t intend to re-use later in the report. Define EVERY math symbol when it first appears in the report.

When citing someone, use their surnames only. Don’t include first names or initials. For example, say “Chua proposed…”, not “L.O. Chua proposed…”. When citing a paper with two authors, include both authors’ names, not only one of the authors. When citing a paper with 3 or more authors, use either et al or include all authors names, for example, Lee et al… or Lee, Lin, and Wang…

Use standard forms of references, for example (must include author names, title, source, page number, year, etc. Note there should a space at appropriate places.),

[1] J. C. Maxwell, A Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73.

[2] Y. Yorozu, M. Hirano, K. Oka, and Y. Tagawa, “Electron spectroscopy studies on magneto-optical media and plastic substrate interface,” IEEE Trans. Industrial Electronics, vol. 2, pp. 740–741, August 1987.

Write simple and short sentences, unless you are absolutely sure that the long sentences you write are clear and correct. Make sure each sentence uses correct grammar. DO NOT use abbreviations like “don’t”, “can’t”.

Each reference must have publication title and year. For a journal reference, make sure it also has journal name, volume number, and page number. For a conference reference, make sure it has conference name. For a book reference, make sure it has publisher name and publisher city.

For each figure plotted:

1. If the width of the figure is about 6cm (about half of A4 page width), the font size of labels (words/numbers) for the axis or inside the figure must be 8 or 9, so that it is big enough to read. The size of the axis label should be increased proportionally, so that when the figure is reduced to 6cm- wide later, the axis label is still big enough to read.

2. Minimize the number of words inside of a figure. Explain the figure in the figure legend.

3. Do not use color in figures. Instead, use different line types (dashes, solid), different dots, squares, triangles, etc. Do not have too many of these symbols in figures to make it hard to tell which is which.

4. Avoid grids inside a figure

5. Do not have too many ticks on each axis (for example, no more than 6 ticks for each axis). E.g., figure 3.1.7.

6. Each figure or table must have a legend, a concise description of the figure/table. Do not use shades or colors in tables.

7. Make sure all figures have high resolution. If you need to capture screen, enlarge the figure first and then capture the screen.

8. Avoid figure titles (labels on top of the figures). Have description in figure legends instead.

All plotted figures should have axis and all axis must have labels (i.e., explain what is plotted against what).