Learn is an interactive application based on a simple artificial learning algorithm. The little blue balls are born, they move according to gestures defined by the user and then die. Click above to open (new tab).
Watch an animation created using this algorithm here (new tab).
Each gesture made by the user is included in the memory of all agents that are present. Each agent behaves independently, moving about using one of the movements learned, chosen by random. New generations of the virtual agents are identified by different hues of blue.
In order to limit the number of simultaneous agents, the balls have a controlled life span. Only the first agent is preserved in order to allow the application to run continuously. The memory of the first agent is continuously cleared (older memories first), also as a way to allow the system to always return to its initial state in case there is no interaction.
This is an updated version of the original experiment, created in 2009. The overall behavior of the system was changed, and some adjustments and aesthetic changes where made, such as:
- Bigger area (fluid, adapts to resolution);
- Interface, with comments and additional information;
- Gestures now produce a “stroke” on screen;
- Various adjustments to variables of the system;
- The Amoreiras Project is also based on artificial learning;
- The game Nibballs uses a similar mechanic (also, balls!)
[This is a translation of the post in portuguese from 2013/12/06.]