Computational Neuroscience – The Second Term

I am already right in the middle of my third term studying Computational Neuroscience and so far I have not got around writing anything about the second semester. In my defense I spent most of the semester break studying for exams. But finally I have the time to do my report. First, you might want to read my post about the first semester if you have not done so already. I will not repeat all the general information. Instead I will start of with the individual courses I took, tell you something about the social events and finish with some general personal remarks.

Courses

Mostly the modules of the previous semester were continued with a change in focus topics. I will start off with these courses and then move on to the new and great Models of Higher Brain Functions before closing with some words about less important courses.

Machine Intelligence

The Machine Intelligence (MI) lecture was continued from the previous semester and we moved on from supervised methods to unsupervised methods. The following topics were covered:

  • Probability Density Estimation
  • Principal Component Analysis (PCA)
  • Independent Component Analysis (ICA)
  • Stochastic Optimization (i.e. Simulated Annealing and Mean-Field Annealing)
  • Clustering methods including K-means Clustering
  • Self-Organizing Maps

Some of the topics were new two me, others I already had covered in my Bachelor’s degree.

The course was held like in the last term by Prof. Obermayer in the same structured, but dry style. However, there were multiple occasions on which our tutor Timm Lochmann did the lecture. The exercise sheets contained definitely more programming exercise than the term before (which was fine with me).

To complete the Master’s degree one has to take a one hour oral exam covering both parts of the lecture, but you can choose the date pretty free. Nevertheless, I would recommend to do it quite early (e.g. some weeks) after the end of lectures. All the other exams of the second term are pretty late in the semester break which gives you plenty of time to prepare. I assume, there will no phase which gives you more time to prepare during the Master program (at least if you try to stay within the two years it should last). Also, the later you take the exam the more you will forget.

So I took the MI exam as the first in this term. Unfortunately, Prof. Obermayer had to go to the hospital and was not available for several weeks (he is better now). Because of that they first moved my exam date and then the exam was not done by himself, but our tutors of the two semesters Timm Lochmann and Wendelin Böhmer. I am not sure whether that was to my advantage. Wendelin knows really a lot and expects you also to know a lot. He also asked stuff from his tutorials whereas Timm said beforehand, the tutorials will not be covered in the exam, and I did not look at them again. Even though, I definitely cannot complain about my grade. :)

Acquisition and Analysis of Neuronal Data

Acquisition and Analysis of Neuronal Data is another lecture distributed over two semesters. After the first more practical part we got more the theory heavy stuff. The first half of the semester Prof. Kempter taught us about different ways to estimate firing rates, quantifying stimuli, measuring correlations, spike train decoding and estimating information rates. Then the second part about Brain Computer Interfaces (BCI) based on event related potentials and motor imagery was done by Prof. Blankertz.

Prof. Kempter’lecture syle was pretty much the same as in Models of Neural Systems in the first semester. Thus, his lectures are okay, but nothing spectacular. I would say mostly the same about Prof. Blankertz with the addition of one thing I liked quite much. He showed us frequently graphical interpretations (e.g. of whitening) which made understanding a lot easier.

In opposite to the first semester, we had to do some programming exercises this semester. The BCI part had to be done in Matlab which is a bit unfortunate as Matlab is not taught in the Master (only Python). In my opinion Python is also way better than Matlab. In the end we also had to do a small project for the BCI part. That was basically just a normal, but longer, exercise.

This course has to be completed also with a one hour exam. It will be done by the four main lecturers and you will be asked half an hour by two of them and then another thirty minutes by the other two.

Models of Higher Brain Functions

The third big course of the semester was Models of Higher Brain Functions covering topics like Slow Feature Analysis, Reinforcement Learning, models of attention, models of decision making and models of multisensory integration. The lecturers were Prof. Sprekeler and Prof. Kiebel and they did an excellent job. Especially Prof. Sprekeler delivered some of the best lectures I ever had.

In addition to the main lectures there was also a block course before the start of the semester about Cognitive Neuroscience by Prof. Haynes which was also quite interesting and in the last weeks of the semester we did a seminar about learning and memory. For this seminar every group of usually two persons had to do a practice talk. For each practice talk Prof. Sprekeler spent an hour and gave valuable feedback. I have rarely seen such a commitment.

Finally, we had to do a 30 minute oral exam with Prof. Haynes and Prof. Sprekeler. Even though I really loved the lectures I somehow got my worst grade in the Master’s degree so far in this exam.

Introduction to Philosophy of Mind

In addition to the mandatory courses I attended a class about Introduction to Philosophy of Mind by Prof. Pauen. I decided to do so because I was interested of the topic and already knew that Prof. Pauen does good lectures. For writing two pages about one of the lectures I got two ungraded credit points which I can use for the module Courses on Advanced Topics.

GRK Lecture Series

I continued with the GRK lecture series. There is not much to say in addition to what I already wrote about it last term. Except that, it took them really long to fix the grading. In the end you were allowed to not hand in three or four exercises, but all handed exercises would have influence on your grade. The result of this was that I did not do the last exercises because they might have had a bad influence on my grade. Also, as a Master student you had to put more effort into it than a PhD student in case you cared about your grade. I think, they will change the grading again this year to fix this a bit.

Individual Studies

As I wrote in the report about the last semester I agreed with my mentor to read several chapters in Principles of Neural Science by E. Kandel. This semester I met with my mentor and we talked for about 45 minutes about it. That way I obtained 2 ungraded credits which I was still missing for the individual studies (4 I already had from the math prep course).

If you also want to do a reading assignment I would suggest to look for a different book. Even though the Kandel explains things well it is quite detailed and a book giving you a broader view might be better suited in the beginning.

Social Life

The social life in the winter at the BCCN has been discussed in the previous report. But in the summer there is not just the Stammtisch, but also barbecues which are quite fun. Even if you happen to be vegetarian as many people at these barbecues are actually vegetarian and grill vegetables and grill cheese besides meat.

Another important social event is the yearly retreat. Mostly Master and PhD students stay for three days somewhere outside of Berlin (this year Schloss Tornow) and each day there is a number of interesting talks and some more or less interesting group work. The main talks this year were done by Idan Segev and Will Penny. In the group work my group had to come up with a PhD proposal for a project which can be done by an average student and results in publications in Nature and Science. This definitely was not the most fun project. But there were others like writing an article about Neuroscience for a boulevard newspaper. The end result was quite funny.

Besides this more academic site we also had some leisure time which was filled by a canoeing tour and excellent meals. Each evening there was, of course, a barbecue (and a bonfire).

Conclusion

After the first term was a bit disappointing for me, I really liked the second term. One important factor was that I reduced the time I spent on my part time job. That made it a lot less stressful. Also the Models of Higher Brain Functions lectures were excellent and finally gave me what I expected from the program. Maybe, I also liked the second term better because it focused more on higher level, more abstract stuff without all the nasty, fine-grained biological detail of neurons.

Unfortunately, the second term ends with a number of big exams which made me spending most of the break with learning. But you get rewarded with having all important exams done and you can look forward to a year of working on real research projects in your lab rotations and Master thesis. :) Actually I already finished my first lab rotation by now, but that is story left to be told at a different time.