„fuck! I built an ambient light“

Aluminium frame, LED panel, Raspberry Pi, video card, processing,  artificial neural network (ANN)


The light installation "fuck! I build an ambient light" is a complex system with self-organising and self-learning behaviour, at the same time it is an artistic study of contemporary deep learning techniques. The installation explores non-hierarchical arrangements and collective behaviour by implementing such a system visually through LED landscape. The interaction of many entities creates a complex visual behaviour. It seems as if they negotiate their position with nearby actors.

The installation consists of 24.576 LED's (128x192px) arranged in a two-dimensional grid of 800x1000mm. 

All LED's are randomly set to color at the beginning so that in each iteration the system finds a new initial form. 
The installation drifts during its runtime through different activation levels and thereby constantly develops new formations and constellations until they find a defined end. When the change in the learning rate decreases to such an extent that no further significant change takes place. 

A self-organizing map (SOM) is a kind of artificial neural network.  It is trained with "unsupervised learning" to produce a low-dimensional (two-dimensional), discretized representation of the input space of the training samples, a so-called map, and is therefore a method for dimensional reduction. Self-organizing maps differ from other artificial neural networks in that they use competitive learning as opposed to error correction learning (e.g. backpropagation with gradient descent) and in the sense that they use a neighborhood function to preserve the topological properties of the input space.