Close on the heels of another bootcamp in Madison, WI, for the past two weeks I had the pleasure of teaching a 30 hour scientific computing course at the African Institute for Mathematical Sciences (AIMS) in Cape Town, South Africa. The course was an expansion and refactor of Software Carpentry material to fit the AIMS need. You can find the repository of source material here.
I could gush about how I fell in love with Cape Town, the unexpected preponderance of vegan cafes in Muizenberg, the sea, the sand, the mountains, and la dolce vita I am otherwise used to from Santa Barbara. I will not. AIMS is about the students. Let's examine them after a brief tour of AIMS itself.
AIMS, in association with its partner universities, provides a one year structured masters degree to exceptional students from all over Africa. This provides a needed stepping stone for many its students to pursue other universities across Africa and across the world that they would otherwise likely not have to opportunity to attend. The social goals at AIMS go hand in hand with the technical mission.
The Cape Town AIMS is the first such center and was opened in 2003. However, other AIMS centers are springing up in various other countries on what seems like an annual basis. There is also an associated AIMS research center in Cape Town which houses some doctoral and post-doctoral students as well as a related center dedicated to improving the educational methods of secondary school teachers (AIMSSEC).
The complex itself is a 4 story, 0 indexed building less than 100 meters from the ocean. Floor 0 contains a lecture hall, offices, kitchen and dining area. Floor 1 has more offices, the library, and a world class computer lab. Floor 2 houses the student dormitories. Floor 3 houses rooms for some staff and visitors, such as myself. In short AIMS is a mathematics monastery straight from my undergraduate physics major dreams.
The view from my room at AIMS, no fooling.
This year's class is approximately 50 students and hail from Egypt to Chad to Nairobi to South Africa itself. In the neighborhood of 20 countries were represented in the student body. As William, a student, put it "by visiting AIMS I visited all of Africa." However, national origin is only where the diversity starts.
40% - 50% of the students were women. Furthermore, It seemed that the age distribution ranged from mid-twenties to mid-forties. Additionally, though the ability to read and understand English was generally quite high, the ability to speak or understand spoken English varied from fluent to novice.
Technical ability was also all over the map. Some of the students it seemed had been professional Java developers in a past life. On the other hand there were some students who had learned to program Pascal out of a book and arriving at AIMS the week before was their first time ever using a computer.
Still, the students themselves were incredibly enthusiastic. Far more so than their Western counterparts I have taught. The computer lab at AIMS is open 24/7. I would often pop in between 11 pm and midnight and categorically I found 20 - 30 students still at their machines, working through the day's material, and discussing what they were learning in small groups.
A cynical person would claim that this phenomenon is due to the students having nothing else to do in sleepy Muizenberg. Naturally, they gravitated to the place where they have internet activity. I claim instead that the students acted with the purpose of learning. No one sits in lectures for six hours a day, day in and day out, in a language they are still acquiring, and then hits the books for another six hours afterwards if they do not feel that what they are doing is important and valuable. Perhaps this mathematical monastery should more rightly be denoted AIMS: Academics Inconsequent of Mortal Stamina.
The Software Stack
Throughout all of AIMS, including the staff and administration, the software distribution is a model for how other, similar centers around the world should operate. Their stack is set up and managed by the preeminent Jan Groenewald who I know through the illustrious Katy Huff. The operating system is Ubuntu, the mail client is GMail, and courses are taught only with open source scientific software such as Python, SciPy, R, Octave, LaTeX, and Sage. This is my stack.
One of the neatest teaching aids that Jan and I ended up test driving and deploying is an application called epoptes. This is basically a screen sharing program with some extra bells and whistles that allow you to take over the student's computer as well. It was incredibly effective as long as you did not accidentally hit the button to shut down all of the student's machines.
If there was any lingering doubt, it has now been scrubbed away. The open scientific software stack is completely ready for prime time in the classroom. Anyone who is trying to sell you something is doing exactly that.
I taught the entire course from a series of IPython notebooks cribbed from Software Carpentry. Some of these were tutorials I had originally written a couple of years ago. Humorously, I found that the documentation lesson had not been significantly altered in that time.
However, forking from Software Carpentry required significant amounts of extra effort in some cases to make the lectures complete for my AIMS needs. (I have since pull requested these updates back though they are waiting for a merge.) I was very tempted to use some material from the Python Bootcamp at UC Berkeley. I did not because of licenses. Software Carpentry material is CC BY v3.0 licensed. The Berkeley Python material is CC BY-NC-SA v3.0 licensed. Even though I am not currently planning on reusing this material, the threat that I would be forbidden from doing so in some circumstances was enough to drive me away from the non-commercial copyleft terms. Ironically, on the first Tuesday I was at AIMS, the class took a field trip to the University of Cape Town to see Richard Stallman speak. He convinced me.
In terms of the instruction, I had the incidental privilege of giving the first lecture of the academic year. It became clear after a few lessons - introduction to scientific computing, BASH, git - that I would have to adapt my teaching style. Luckily, the students shared an advanced mathematics background. I could assume that they all knew linear algebra, differential equations, and set theory.
Teaching to the middle 80% at AIMS would not have worked. The students were too diverse of a group in too many ways. Focusing on the mythical middle 80% would have meant leaving everyone behind in some way all of the time rather than leaving some people out most of the time. I felt that if I had chosen to teach to the middle 80% I would be ignoring the core missions of both AIMS and Software Carpentry.
So what changed? Linguistically, I tried to teach to the bottom 10%. The students who already knew English did not mind that I spoke slowly and directly and repeated myself often and it made a world of difference to those who did not know English as well. I did not always hit my 10% goal but it was certainly worth trying to get there.
At the tutors' (teaching assistants) request, we also instituted brief reviews of previous material. Since I taught the shell on the first day we went over useful commands on second and third. This gradually morphed into a generic question and answer session at the beginning of every class that would last up to 30 minutes of a two hour block. For many students, the multi-day reinforcement was critical.
That said, the most fundamental change was that I taught as much as I could in analogy to mathematical principles. Functions are functions, operators are operators, and sets are sets. This made certain topics, such as slicing and recursion much easier, but other topics, such as decorators and scoping, a little more difficult.
Overall, after I made these adjustments, the course went smoothly. I would love to have the opportunity to go back and restructure the entire course with AIMS students in mind now that I know who they are and how they differ from the standard research-distracted Western graduate student.
As any good instructor would do, I ran assessment surveys before and after the course. This is as much to gauge the students as my teaching ability. Since the course was derived from Software Carpentry I used the official pre- and post-assessment surveys. Since the raw data is not anonymized, if you would like access to it please contact me.
On the subject of programming to do basic data analysis tasks, a huge leap was made over the course of the two weeks. Figure 1 shows a comparison for tasks involving opening a plaintext file, reading in data, and computing summary statistics. Most students went from not knowing where to start to feeling that they could successfully accomplish this task given enough time.
Figure 1: (a) Pre-assessment question, "Write a short program to read a file containing columns of numbers separated by commas, average the non-negative values in the second and fifth columns, and print the results." (b) Post-assessment question, "A comma-delimited file has two columns showing the date and the number of sun spots recorded for that day. Write a program to produce a graph showing the average number of sun spots for each month."
However, the most staggering advancement was with respect to the shell. This makes a certain amount of sense as BASH was the first technical subject we discussed. Consequently, the shell was the most reinforced material. It was reviewed at least once per day. Figure 2 shows that the distribution in ability basically flipped.
Figure 2: (a) Pre-assessment question, "In a directory with 1000 text files, create a list of all files that contain the word Drosophila, and redirect the output to a file called results.txt." (b) Post-assessment question, "How familiar do you think you are with the command line shell?"
Lastly, I made an unorthodox decision to try to teach version control as early as possible. So the second day, right after the shell lectures, we spent 4 hours on the details of how to use git. I mostly received blank stares. However since I ran the course from a bitbucket repository the students had ample opportunity to reinforce what they learned. Figure 3 displays an advancement that, while not quite as staggering as those above, is great to see given the importance of the topic. I doubt that any of the students would now say that they do not know what version control is and what it is useful for, even if they claim to not know how to use git.
Figure 3: (a) Pre-assessment question, "Check out a working copy of a project from version control, add a file called paper.txt, and commit the change." (b) Post-assessment question, "How familiar do you think you are with the version control stem used in your bootcamp?"
Overall, I find these results very encouraging. I was able to seed some fundamental best practices. However, it is now up to the AIMS staff and tutors to shore up these ideas with continued education and practice. As this was a self assessment, and not a competency assessment per-say, the students definitely seemed to take something away themselves.
At the end of the two weeks, the students sat me down in a chair at the front of the class and presented me with a thank you card. Along with this they read a statement prepared by the class. This was definitely one of the most rewarding moments of the course - and possibly of my life to date. I wanted to laugh and cry at the same time. The statement went as follows:
We are so grateful for teaching us the course Scientific Computing, the Python language especially. We are also grateful to your style of lecturing in that you welcomed questions and you were so patient with everyone despite our variety of accents. You answers every question by first saying, and I quote "Aaaah, Okaaaay, that's a very good question." This encouraged all of us to ask question and we benefited - we are really grateful.
Taken while I listen to the goodbye speech from the students.
The student William from before also got up and said a few words, some of which were, "Anthony, you are not a trained teacher, but a born teacher." I really appreciate this comment on all levels.
However, the real showcase of the impact that this course had is in what the students wrote in the card. My favorite and most telling inscriptions were:
Audrey = "thank you Anthony"
keep in touch - git pull
I may forget Python, but not you. Thank you!!
Personally, I would like to thank Barry Green, Jeff Sanders, and Jan Groenewald for giving AIMS its superior environment and extending to me the opportunity to come and lecture about some of my favorite topics. Additionally, I'd like to thanks the class tutors, Adriaan van Vuuren, Antonio Cardoso, Eva Ujeneza, Evans Ocansey, Markus Kruger, Mekdes Tessema, and Morgan Magloire Pene, there is no doubt in my mind that the class would not have run half as well.
In closing, AIMS is an incredibly special place. If you ever have and opportunity to visit the Cape Town campus or another location I highly encourage you to take it. I promise that you will not regret it!
Jubilant photograph with the students after the final lecture.