Ph.D. Seminar Presentation

Ph.D. Seminar is a mandatory 4-credit course required for getting through the Ph.D qualifier. A grade of minimum 8 is required for qualification. Title of my seminar is Construction of Gene Networks from Expression Profiles. Seminar is mostly a literature review of the topic. Structure learning had been of my interest for quite a while, and I thought I would be wetting my feet with a review of Gene networks construction. I discuss here mostly about the preparation and soft-skills.

Report Writing

The task of report writing is to make a summary of the papers read. I read about 16 papers on the topic. I set myself a target of one week to write a report. I even told my guide that I will send him the copy of the report within a week. Having a hard deadline helps me focus. It took 2 days longer.

For writing the report, I first started in the analog form. I wrote the notes of each paper on a single sheet of paper - A4 sheets written with a HB pencil, and pasted all the 16 sheets on the wall. When writing in a single sheet, I was trying to answer 'What did I learn from the paper that I can summarize in a single sheet ?'. I didn't write the whole thing not without referring the report (though I should have). Then, I moved around these sheets, changing the order, until I got the order and the topics right. I found that this exercise helped get the over-all picture. I think getting the big-picture is a very important step in writing a decent report. It helps build a narrative.

I used XeLaTeX to format the report. With XeLaTeX, one can use open-type fonts. For serif text, I use, what I think is the most readable typeface, the 'Minion Pro'. It is my favorite type. Many science journals including 'Nature' use Minion. For sans-serif text, I use the 'Myriad Pro' type. (Steve jobs uses Myriad typeface in his keynote talks. Its Apple Inc.'s corporate type). Both these fonts are available, if you install Adobe Acrobat Reader. (If I can, I would buy the complete Adobe Font Folio collection for 2600 USD)

LaTeX formatting for the report was inspired from uggedal's thesis. For figures in LaTeX, I used tikz. I initially thought, I would use vector drawing software such as ink scape. But I will find any excuse to do programming. I choose TikZ. TikZ, with its clean interface, is flexible and much easier to draw than MetaPost (I don't know anything about pstricks). Source code for my report and its figures is here (Source added only to use it as an example. No guarantee that it would even compile).

In hindsight, the problems I find in my report (and its making) are

1. Too elaborate introduction : Since my topic is biology, and I assumed a B.Tech CS student to be the representative of my audience (which normally the advice given M.Tech seminars), I had to write a whole chapter of 'Biochemistry 101'. It just increased the number of pages.
2. Too much time spend on formatting the report : As some one who would rather program than write, often my programmer devil sometimes emerged victorious against my writer angel. That lead to spending too much time figuring out how to write my latex macros, TikZ diagrams etc., and less than optimal time re-writing. I have to remember this : "All good writing is re-writing".
3. Elaborate explanations : In the end, I found the report too bloated. I gave to a friend to read and give me feedback, and she told me that the report is just too big. The difficulty with writing seminar report is the problem of telling Ramayana. They say, "You can tell the whole story of Ramayana for years and years. You can also tell Ramayana in the time one throws and catches a lemon (Lemon goes up - Rama shot. Lemon comes down - Ravana died. Did I spoil the ending for you ?)" The problem is striking a balance for the audience. I hope I get better with time.

Seminar Preparation

I used Beamer to build my presentations. Since I had used TikZ to write the report, it is easier to do simple animations with Beamer. Simple animations are useful in building concepts. Instead of starting with a complicated picture on a slide, one can build the picture frame by frame - just adding parts of the complicated picture each frame. Each part should be independent enough that they can be explained separately, and each part has to be related only to those parts already in the slide.

My presentation slides are here. This is not the first version. The first version was much longer and much bigger. When I rehearsed the first time, it took 80 minutes! I am only allowed 20 minutes. I had to rehearse four times to get down to 20 minutes. I recorded each presentation using the screen-cast software Camtasia. It is a useful for catching the mistakes.

In hindsight, the problems I find in my preparation are:

1. Too big : It is big. I thought I could squeeze in every thing I wanted to say. One always over-estimates their ability - especially when one just starts learning. What I should have done is that I reduce what I have to say, rather than trying to squeeze more. The most important skill one needs is the guts to remove stuff. As Antoine de Saint Exupery said, " Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away."
2. Too much focus on the visuals: I spent quite a while on how the slide should look like and how each picture should assemble. I realize the time could have been useful in editing my content, and stream-lining it.
3. The voice : My voice has the monotonous frequency, which I think has a pretty good chance of tearing the examiner's ear drum upon the chance that it is natural frequency of the ear drum. I need to develop a rhetoric rhythm for my modulation - some sort of radio voice. Should I apply for a part-time acting course ?

Seminar Presentation

What happened in the presentation is that my audience changed, and I couldn't adapt myself. I prepared my presentation with an audience of a B.Tech Computer Science student, but what I had was only my examiner Prof. Sunita Sarawagi. I can't blame anyone but myself. I knew she was going to be my examiner. Since I wrote the report with a student audience in mind, I continued to prepare my presentation so - being truthful to the report. She was much more interested in the Machine Learning aspects of the problem (I should have known, I am enrolled in her Machine Learning course).

Another trouble I had was that I was bogged down with a question I don't know the answer. I managed with an improvised answer. But when I later thought about it, I had a much better answer. I have to learn to reduce this lead time of the thought process between Q and A.

Other than that, the presentation went fine (for some definition of 'fine').