Emergent Program Synthesis Based on Reinforsment Learning and Computer Vision Techniques
Alexey Ruslanovich Pitikin, A.G. Sherstova · 2024
At the moment, the number of developed programmes is growing rapidly all over the world. There is a huge number of development methodologies, but the problem of architecture of complex systems remains relevant. This paper proposes a new approach to software development, which is based on emergent synthesis of system components based on reinforcement learning, using computer vision technologies. The methodological basis of the research is multi-agent reinforcement learning. To demonstrate the work of the approach, a method of interaction of system components is used, based on modeling full-fledged images based on the shapes they consist of. The formalization of the emergent properties of selforganizing systems, as well as their application for software synthesis, is proposed.