Attuning Adaptation Rules via a Rule-Specific Neural Network

There have been a number of approaches to employing neural networks (NNs) in self-adaptive systems; in many cases, generic NNs/deep learning are utilized for this purpose. When this approach is to be applied to improve an adaptation process initially driven by logical adaptation rules, the problem is that (1) these rules represent a significant and tested body of domain knowledge, which may be lost if they are replaced by an NN, and (2) the learning process is inherently demanding given the black-box nature and the number of weights in generic NNs to be trained. In this paper, we introduce the rule-specific Neural Network (rsNN) method that makes it possible to transform the guard of an adaptation rule into an rsNN, the composition of which is driven by the structure of the logical predicates in the guard. Our experiments confirmed that the black box effect is eliminated, the number of weights is significantly reduced, and much faster learning is achieved while the accuracy is preserved.