On an alternative approach to Evolutionary Programming

The field of Genetic Programming (GP) is a popular approach to automatic programming. However, it is plagued by many issues which prevent it from being scalable. These issues include bloat, need for custom search spaces etc. We propose design for a system called Evolutionary Programming System or EPSys, a system in which we try to eliminate many problems plaguing Genetic Programming and hence preventing it from being scaled and being applied to larger real world problems. We then follow it up by a minimal implementation in which we solve a very simple problem. We then prove that this implementation is equivalent to GP. We also prove that this system is immune to bloat problem which has plagued GP since its inception. We also conclude that the current implementation is very much slower than GP due to huge overhead. We then propose some basic improvements to EPSys which could potentially make it the far more scalable than GP.