Anders' law: The longer you have spent trying to solve a particular problem in a programming language, the more likely it is that this programming language already has a built-in function to resolve just this problem.
In addition the built-in function is likely to be devilishly hard to find, but ridiculously easy to locate once someone have shown it to you.

Edit: I've done a few changes to the site today. Only one of which should be particularly visible for a few of you. The @font-face font I was using did not render at all well under Windows so I've change the font. I'm now using DroidSans and DroidSerif.
The other change is that I've stopped using the, in my eyes, unnecessary prefix www on all sections on the site. So now all links to my site should read http://andershoff.net instead of http://www.andershoff.net. Both should still work though.

Blaarghh.

I'm having my last exam this semester in 8 hours.

Sorry about the Norwegian. The image refers to the European Union's Data Retention Directive/Datalagringsdirektiv currently up for discussion in Norway.

If you understand Norwegian and want to know more you can start at stoppdld.no.

Yesterday I went to this talk with Ingrid Daubechies. In (very) short the talk was about the use if image analysis to collect information about paintings. For instance whether it would be possible to separate a fake Van Gogh painting from a bunch of original ones.

In addition she gave a short introduction to the principles used in the jpeg2000 compression algorithm, as these principles share quite a lot with the theory used to do the previously mentioned image analysis.

The above image is a very simplified illustration of the first level of such a compression made in MATLAB. Basically what you see is the average of squares of four and four pixels from the original image in the top left image. As well as the difference from the average to the top right, lower left and bottom right pixels in the same square. Each of these difference can be seen in the corresponding image with a black background. The white lines in the three black images represent areas of the original image where there is a big difference in contrast. In short the lower left image will to a greater degree represent vertical changes in contrast, whereas the bottom right will indicate diagonal changes in contrast and similarly for the top right image only with horizontal changes in contrast instead.

These four images contain almost enough information to reconstruct the original image with the same total resolution as the four images above. It does not, however, require as much storage, as the black areas in the three black images does not need to be stored. If this process is repeated an adequate number of times you will get a (in my case) simple version of the jpeg2000 compression. The original algorithm is quite a lot more sophisticated off course.

Also I should probably note that the above image is not the best image to be stored in jpeg format as the lossless png format for instance still will require a lot less space without losing any information in the process. The idea is still illustrated nicely though.

A few examples of how not to chair.

And I survived Brazil by the way. Some pictures can be seen here.

Only a few days until I'm leaving for Brazil. That is all.

I wonder what went through their minds before they started marketing a dishwasher detergent under the name Finish Quantum?

- That is the question.