June Gloom

Sometimes, one writes computer code to answer a hopefully important science question. In other occasions, code is developed with some economic interest in mind, legitimate or otherwise. And then there are situations where something else is going on. The work shown here, which I call June Gloom, is just that: something else. Specifically, it is about producing pretty pictures. I am not pretending that there is any deeper meaning to this, but I will say that I was inspired by the work of Bridget Riley. The pictures are produced by coloring tiles according to the values of an underlying Potts model. A simple Monte Carlo simulation is employed to generate new pictures from a random initial state. Depending on the temperature setting, as well as some further details of the interaction between tiles, we can produce a variety of patterns.

We start with a random pattern. Every tile has one of six possible colors, selected with equal probability. [pdf] [high-res png]

Is is getting more exciting: turning interactions on and running the simulation for some time, we get something rather different. Tiles of the same color tend to stick together. [pdf] [high-res png]

Finally, we change the interaction between the tiles. Now, like-colored tiles do not want to be close to each other any more. The pattern looks quite random again, even though it is different from the first one. [pdf] [high-res png]

The code is available on Github. The colors are from ColorBrewer 2.0, which I can really not recommend enough if you are looking for a nice color scheme.