Machine Learning

XKCD has a wonderful new comic: These days there also was an article in MIT technology review criticising the same issue. For me this is a daily issue. My colleagues usually are happy to just get a quick test procedure; usually on smallish data. Chances are it pattern matches some noise. But nobody will find out as reading what a classifier found is near impossible. I mean sure we weed out all the nonsensical predictor variables – but even on the sensibel there is additional information mixed in.

Nice stuff from uibk

As I was checking whether the uibk would offer us some storage for git projects I was positively surprised they run a full gitlab instance. Unfortunately we will be limited to 50 projects - so I will not be able to abandon my bitbucket account. Anyway nice to be able to host potentially sensitive stuff internal. Also they run mattermost - now that looks very interesting. Chatting without going through american services… Need to get my friends in there.

Configuring offlineimap

Having most of my workflow on the commandline mailing my results using Tunderbird or is sometimes quite a chore. I would rather do it from inside my workflow. mutt seems like a natural option. I also quite quickly considered alternatives like alpine - though since I prefer vim for my text editing I quickly decided against it. offlineimap For mutt it quickly became aparent that I would need some external tool to fetch my mail.

Story on Moebius

Vi Hart tells this lovely mathematical story.

Publishing work flow template

After some time my cookiecutter templates for my data analysis work flow is finally in a state that I can show the world. You can get it from github. All in all it is my way to setup a reproducible research work flow - for too long I have just copied files around. It is all about using external tools in addition to R. The most essential being make. I am very proud about the makefile that keeps code and data directories clean.

ERJ accepted correspondence

Nice the ERJ has accepted our correspondence about the future of respiratory medicine. Hope I will have the link soon.

Unfortunately I had extremely limited room to address the statistical issues. Anyway writing this correspondence was a pleasant experience. Really hope we will have more to say soon.

So relatable.

This is just soo eerily familiar. comic phd comic

Ever wondered about statistical mechanics?

At least the book got it right 😺

Publication word cloud

I wanted to play with word clouds for some time. Usually the problem is to obtain a nice corpus. Luckily PubMedWordCloud was recently updated and I had something to play with. The initial cloud looks not good enough - I did not like how plurals are counted separately. Since I couldn’t clean it up easy enough I ended up going back to the tm package and roll my own script. Unfortunately it’s far from publishable.

After three months finally an Euler result again

Having spent three months on and off a problem I finally managed to sufficiently solve Problem 110. Strange how problems change once you work on huge numbers.

But now I can again proudly present my score Project Euler
Score. Nicely this puts me in the statistics of active ones on place 19 in Austria.

Diese Seiten werden auf einem Server der Universitaet Innsbruck gehostet und unterstehen diesen Richtlinien.