Integrating Gaussian Process with Reinforcement Learning for Adaptive Service Composition

Service composition offers a powerful software paradigm to build complex and value-added applications by exploiting a service oriented architecture. However, the frequent changes in the internal and external environment demand adaptiveness of a composition solution. Meanwhile, the increasingly complex user requirements and the rapid growth of the composition space give rise to the scalability issue. To address these key challenges, we propose a new service composition scheme, integrating gaussian process with reinforcement learning for adaptive service composition. It uses kernel function approximation to predict the distribution of the objective function value with strong communication skills and generalization ability based on an off-policy Q-learning algorithm. The experimental results demonstrate that our method clearly outperforms the standard Q-learning solution for service composition.