Presentation day

※ This article discusses the 4th of October 2016.

On Tuesday, the 4th of October, I was asked to give a presentation on the introductory chapter of Bishop’s book on Machine Learning and attended my first lecture.

After a night without sleep, you feel fresh and awake in the early morning. This changes a lot during the day. I took a shower, ate some breakfast and waited for Martina to meet up. This helped me to reduce my anger, which developed during the night. The anger is mainly due to a lack of a good internet connection. So most of the time during the night I didn’t have internet access, which annoyed me a lot considering my todo list. Tasks on the list mainly depended on internet access, therefore I could only do the most important stuff with a lot of patience: The presentation and finding lectures, I want to attend.

I want to separate this blog post by topics.

## Syllabus

We got to university the usual way; by train and walking up the hill by foot. At university, I attended my first lecture; Japanese on elementary level. Yes, it was necessary to decide upon the classes I take.

Technically, I prepared very well for the selection of courses. The course search interface at Kobe University is terrible. At TU Graz, the interface has little features, but if you plug in English course names, it will find the course. At Kobe University this is not guaranteed (even if you copy&paste the name). In May I wrote a python program which automatically retrieved all the university courses and stored them in a SQLite database. I provided HTML templates to dump filtered outputs of the database. This helped a lot to handle prerecognitions.

Prerecognitions are certificate which certify (by deans of a curriculum) that after successfully completion of a course, it will be recognized for the regular curriculum. So I used my script to retrieve course IDs, titles, lecturers and descriptions. This was essential to be able to type all course information into CampusOnline quickly (for TU Graz or Uni Graz respectively). I pre-recognized all possible combinations like the same course with two different IDs, on different days or sessions or the "semester"-combination as well as the "quarters"-combination. Kobe University is currently transitioning from a semester-based schedule to a quarter-based schedule.

Well, it became very awkward, when we recognized CampusOnline shows the original course names (therefore Japanese), but the print server does not have proper JCK typefaces installed. Therefore CampusOnline prints only showed lots of rectangles for uninterpretable characters and only the course ID was left. With this ridiculous paper, we showed up at the dean’s office. Luckily enough, all deans have been lenient.

And now I got very annoyed by this whole bureaucracy, when in September 2016 Kobe University replaced its search interface and also its course syllabus. My script did not work any more and all courses got updated. Every course ID got longer (typically it got two additional characters as prefix) and the script failed, because Kobe University now tracks you with a search token. You cannot parameterize any HTTP request, but a unique token will be updated for every POST request you do.

In conclusion:

  1. WiFi is too unstable to look up all the new course data for now.

  2. Tomorrow morning I want to attend classes, which technically started today.

  3. My program to retrieve data does not work any longer anyway.

  4. And many courses I intended to attend (and therefore prerecognized) don’t even exist anymore.

I was terribly frustrated this night and simply concluded: "Screw you! I will take Japanese classes, do my bachelor thesis and don’t care about anything else". So I want to do the intensive Japanese class at elementary level. I will proceed with my bachelor thesis as planned. And I will probably run into troubles with recognizing probability theory and statistics, because I am not able to attend classes, because of the intensive Japanese course.

I will lay out my schedule in a different blog post.

In class, we were two people. Two Austrians already familiar with Hiragana and Katakana. So our teacher sped up a lot with the content and revised basics of Japanese. In the end the lecture was a very good revision and it seemed a lot like individual support. However, being tired makes it difficult to follow the lecture and also I want to take the Intensive Japanese class in the future, not the regular one.

I also remembered taking a photo of the temporary WiFi password to fix this issue.

## Preparing for the presentation

Last night, I read through most of the chapter. I am not familiar with Bayesian networks and therefore could go through that content, but the general outline of Machine Learning is still in my mind. I had a lecture on this at TU Graz in 2013. For the presentation I learned (by my tutor) that projectors are not often used and blackboard are used on a regular basis. Therefore I prepared without requiring a projector and wrote down what I will write on the blackboard. During the day I needed to finish a program illustrating the problem of polynomial curve fitting.

So after Japanese class, I headed to the room where I am allowed to work at and put my stuff there. I worked on my program. Because I was so tired, I fell asleep multiple times, because it is so quiet. Anyways, writing it in python took me little time and I finished before Lunch time. At lunch we met Martina’s friend Choko, we already know from Austria. She showed us how to proceed at the cafeteria. After lunch I got back and tried to port my program to Lua. After lunch, concentrating was especially difficult and I decided to take a nap like Japanese people often do. I dozed off for about 15 minutes.

After some time after I woke up, a secretary came in (I knew her from the day before when enrolling) and asked me to write down my name. I did and wondered what this is about. Apparently, I attracted someone’s attention and they checked whether I was allowed to be in that room. At this moment I did not feel welcome anymore.

Anyways, I focused on my presentation and couldn’t finish the last ~20 lines of Lua code to finish the Lua port.

## Presentation

In time, I went to the presentation room. I introduced myself to the audience of 4 graduates or PhDs (I guess) and, of course, my advisor. I started my presentation and had a nice line in my presentation even though I got stuck sometimes on minor issues (distinction between terminoloy, etc). I think my flow was nice. Then it was pointed out that 10 minutes were left and I tried to skip content to introduce the Curve Fitting Problem. I became very chaotic and didn’t know which parts of the presentation I can (not) skip. Anyways, time was running up and my professor asked me whether I know certain fields of research. It is really sad, that I have not attended a course on measure theory yet.

So my presentation was fine (I guess) and I stayed to watch my tutor at his presentation on matrix factorization in Japanese. I recognized that their style of presentation is much different from ours. I mean I generally stick to a style I primarily learned from American universities through Massive Open Online Courses, because Europeans present in an unstructured way too often. The best way to describe this is that lecturing in Japan requires a much less confrontative style which avoids interaction with the audience to some extent. Furthermore the goal of my talk was to show that I understood the content and want to present the subject to the audience. However, as pointed out in conversations after my talk, I should raise questions, I don’t understand and skip everything which is understandable. This approach seems interesting and it will take some time to adapt to it. I have my next presentation in 1 week.

I asked my advisor to sign me up for the intensive course. He asked for an application sheet. I wasn’t aware of any. Martina remembered the very first Excel sheet, we got. Yes, it was the right one and I don’t need to find a printer. Sending it per email and I was done for today.

Afterwards I was completely wracked. I was never at home that fast, because I dozed off in the train. Luckily Martina helped me out to find the right way. In the end, I fell into bed at about 18:30 and Martina headed to the supermarket. Thank you, Martina!