Emergent Distributed Systems: self-assembling and self-writing software without the programmer

新兴分布式系统：无需程序员即可自组装和自编写软件

• 批准号: 2143887
• 金额: --
• 依托单位: Lancaster University
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2018
• 资助国家: 英国
• 起止时间: 2018 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2143887
• 关键词: Emergent; Distributed; Systems; self; assembling

The proposed research intends to explore the application of Transfer Learning to the Combinatorial Search Space found when composing individual behaviours (components), through applying a Transfer Learning technique known as Code Fragments to the Combinatorial Search Space. Providing a novel approach to reducing the effect re-computing previously learned combinations may have on a changing set of requirements, an issue encountered by search and optimisation techniques used to identify optimal compositions. In addition, to further exploring the discovery of an optimal distributed (decentralised) solution to a given set of requirements through the placement and migration of Code Fragments or compositions, another open issue within emergent software. Finally, the exploration of fragmenting individual behaviours of components from their compositions is proposed to explore the deployment of highly interactable components onto the same local system has on overall network latency and computation time, with the aim of improving efficiency within the Distributed System.

本研究旨在通过将一种称为代码片段的迁移学习技术应用于组合搜索空间，探索将迁移学习应用于组合个体行为（组件）时发现的组合搜索空间。提供一种新的方法来减少重新计算先前学习的组合可能对不断变化的需求集产生的影响，这是用于识别最佳组合的搜索和优化技术遇到的问题。此外，通过放置和迁移代码片段或组合，进一步探索对给定需求集的最佳分布式（分散）解决方案的发现，这是紧急软件中的另一个开放问题。最后，提出了将组件的单个行为从其组合中分离出来的探索，以探索将高度可交互的组件部署到同一本地系统中对整体网络延迟和计算时间的影响，目的是提高分布式系统内的效率。