synthScape is a machine learning engine designed to evolve executable programs using evolutionary computation in a distributed environment.
Originally, it was created to answer some basic research questions about the relationship between interaction mechanisms, population diversity, and performance efficiency in evolutionary multi-agent systems solving problems in a sequential task domain.
While answering these questions, the core engine turned out to be general purpose enough to evolve any executable programs as long as: (1) a virtual processor architecture is defined, (2) the instruction set for the processor is defined, (3) a simulator is provided that executes instructions on the virtual processor, and finally (4) a feedback function is defined that is able to evaluate the evolved programs to meet one’s goal.
Once the above are clearly defined, the engine can be scheduled to run in a distributed environment, and over time, programs that meet goals start evolving. At the end of the day, it relies on the classic Darwinian Evolutionary process.