A Web Service Composition Method Based on Deep Reinforcement Learning

Abstract

The invention discloses a web service combination method based on deep reinforcement learning. Aiming at the problems of long time consumption, poor flexibility and unsatisfactory combination results of the traditional service combination method in the face of large-scale service scenarios, the deep reinforcement learning technology and heuristic Ideas applied to the problem of service composition. In addition, considering the characteristics of partial observability of the real environment, the present invention transforms the service composition process into a partially observable Markov decision process (Partially-Observable Markov Decision Process, POMDP), and utilizes a recurrent neural network to solve the problem of POMDP , so that the method can still show high efficiency in the face of the "curse of dimensionality" challenge. The method of the present invention can effectively improve the solution speed, and independently adapt to the dynamic service composition environment on the basis of ensuring the quality of the service composition scheme, effectively improving the efficiency and self-adaptability of the service composition in the large-scale dynamic service composition scene and flexibility.

Description

technical field [0001] The invention belongs to the field of computers, and in particular relates to a Web service combination method based on deep reinforcement learning. Background technique [0002] With the continuous development of network information technology, users' functional requirements for software systems are becoming more and more diverse, complex and changing. This trend has given rise to a new concept of design or architecture of software products: Service-Oriented Architecture (SOA), which requires that applications must be developed as an independent collection of interactive services that can provide a good interface. As a new platform for building interoperable distributed applications, Web services are network-based, distributed, self-describing, and modular components that perform specific tasks and follow certain technical specifications. Publish, locate and invoke on the Internet, thus becoming the most promising technical means to realize the SOA a...

AI Technical Summary

Problems solved by technology

However, in the actual environment, the agent cannot fully perceive the environmental information. This kind of partial perception problem belongs to the non-Markovian environment. If it is not processed for the reinforcement learning algorithm, this kind of learning will not be able to converge.

(2) The theory of reinforcement learning focuses on small-scale and discrete problems, but in the real service composition problem, the scale of the service composition problem is not to be underestimated and the states are continuous

The recurrent neural network (RNN) is suitable for serialized data and can simulate these data more accurately. It records the activation value at each moment, and enhances the time correlation of the network by adding a self-connected hidden layer at cross-domain time points. , but on the contrary, this also makes it difficult to train the entire network and the phenomenon of gradient disappearance

Images