This is a research entry associated with a third-party research project or paper. We are not responsible for the contents of any files associated with this submission, or for the accuracy of any code / results. Any questions should be directed to the author(s) of the work.
Distributed Emergent Software: Assembling, Perceiving and Learning Systems at Scale
Abstract: Emergent software systems take a reward signal, an environment signal, and a collection of possible behavioural compositions implementing the system logic in a variety of ways, to learn in real-time how to best assemble a system. This reduces the burden of complexity in systems building by making human programmers responsible only for developing potential building blocks while the system determines how best to use them in its deployment conditions -- with no architectural models or training regimes. Emergent software has been demonstrated to deliver strong results in single, local software systems. In this paper we generalise the approach to distributed systems, to demonstrate how a single reward signal can form the basis of complex decision making about \textit{how} to compose the software running on each host machine, \textit{where} to place each sub-unit of software, and \textit{how many} instances of each sub-unit should be created. We provide an overview of the necessary system mechanics to support this concept, and we discuss the key challenges in machine learning needed to realise the concept. We describe our current implementation in datacentre and pervasive computing environments, and provide experimental results using a base learning approach.
Status: Published. You can share this link.
Venue: IEEE SASO 2019.
Requirements: Dana v226 download
Download: saso2019porter.zip