Inferring P systems from their computing steps: An evolutionary approach

Inferring the structure and operation of a computing model, given some observations of its behavior, is in general a desirable but daunting task. In this paper we try to solve a constrained version of this problem. We consider a P system Pi with active membranes and using cooperative rewriting, communication, and division rules and a collection of pairs of its consecutive configurations. Then, we feed this collection of configurations as input to a (mu+lambda) evolutionary algorithm that evolves a population of (initially random) P systems, each with its own rules, with the aim of obtaining an individual that approximates Pi as well as possible. We discuss the results obtained on different benchmark problems, designed to test the ability to infer cooperative rewriting, communication, and membrane division rules. We will also provide a description of how fitness results are influenced by different setting of the hyperparameters of the evolutionary algorithm. The results show that the proposed approach is able to find correct solutions for small problems, and it is a promising research direction for the automatic synthesis of P systems.

Automated summary

Inferring the structure and operation of a computing model from its behavior is a challenging task. This paper proposes a constrained version of this problem and applies an evolutionary algorithm to find an individual that approximates the original model. The results show that the proposed approach is promising for the automatic synthesis of P systems.