A system learns to draw independently.
An artificial neural net generates self-observed patterns. The ensembler consists of an artificial neural network, a drawing machine and a camera, a recursive system is created. The generative process of learning within the neural network draws patterns. These patterns are a visualization of the learning process of the neural network.
The system influences each other on three levels. One layer processes the image of the camera. The second level, an artificial neural network, learns from the input of the camera. The third layer sends the signals from the network to the drawing machine and the drawing is again observed by the camera. This results in a recursive system that learns to draw patterns on its own.
The neural network, based on the method of the Perceptron network of Frank Rosenblatt and the Hebb learning rule from the beginning of research on artificial intelligence in the 1950s. Both methods still form the conceptual basis of artificial neural networks today.
Artificial neural networks will increasingly operate in our society. They will be able to efficiently evaluate decisions, evaluate data - categorize, recognize objects and learn by data. The project "Neuronal Signs" visualizes the basic structure of learning in neural algorithms.
The project was developed in cooperation with Christian Faubel at the KHM, Cologne.