EPSRC Centre for Doctoral Training in Distributed Algorithms: the what, how and where of next-generation data science

这项CDT将培训60名学生，使他们拥有技能和经验，使他们能够成为分布式算法的领导者：利用“未来计算系统”走向“数据驱动的未来”。商品数据科学已经无处不在。这激发了当今对训练有素的数据科学家的迫切需求。这一CDT将为未来的数据科学领导者提供动力。英国(和世界)需要数据科学家，他们能够最好地利用未来的计算资源来收获新的“石油”：存在于数据中的信息。随着我们毕业生职业生涯的进步，许多核心架构将变得越来越常见。我们预计未来台式机的核心数量将比今天多数百万个。这一核心数量将挑战当前大数据中间件(例如Spark和TensorFlow)所做的假设，即未来计算系统的细节可以与数据科学工具和技术的发展脱钩。更具体地说，数据科学家必须了解如何设计能够在数据移动是关键性能瓶颈的环境中有效运行的算法。为了满足这一需求，我们将提供培训，以确保我们培养出既了解未来计算机硬件设计，又了解如何以及何时灵活使用算法解决方案以最好地利用未来存在的计算资源的高就业人才。从一开始，学生将被嵌入到一个计算环境中，该环境预计他们毕业后将到达他们的办公桌上的硬件资源，而不是今天存在的硬件。这批学生提供了与国际领先的超级计算中心接触的关键群体：STFC的哈特里中心是团队中不可或缺的一部分；我们与美国IBM Research建立的联系将为学生提供使用最先进的计算硬件的机会。这种对未来计算能力的预期将确保我们的毕业生具有很高的就业能力，但也有助于激励最终用户组织参与CDT。我们已经确定了这样的最终用户组织，它们跨越两个主题：国防和安全；制造业。这些主题的组织分别受到绩效要求和效率要求的驱动。我们将根据群体、主题和个人的需求提供培训。每个学生将有两名学术导师(一名与“未来计算系统”相一致，一名与“迈向数据驱动的未来”相一致)，以及至少一名来自项目合作伙伴的导师。这个督导小组将共同界定每个学生的范围。一旦选择并录取了高素质的学生，我们将与学生一起确定符合他们的需求和学生的具体要求的培训。我们提供的培训将包括与“未来计算系统”和“迈向数据驱动的未来”优先领域有关的培训需求。例如，我们将使用来自IBM(用于培训快速通道公务员)和加州大学伯克利分校的客座讲座，以确保我们最大限度地提高毕业生茁壮成长的能力，并成为未来分布式算法的领导者。

This CDT will train a cohort of 60 students to have the skills and experience that enables them to become leaders in Distributed Algorithms: capitalising on "Future Computing Systems" to move "Towards a Data-Driven Future".Commodity Data Science is already pervasive. This motivates today's pressing need for highly-trained data scientists. This CDT will empower tomorrow's leaders of data science. The UK (and world) needs data scientists that can best exploit tomorrow's computational resources to harvest the new 'oil': the information present in data.As our graduates' careers progress, many cored architectures will become increasingly commonplace. We anticipate millions more cores in tomorrow's desktops than today's. This core count will challenge the assumption made by current Big Data middleware (e.g., Spark and TensorFlow) that the details of future computing systems can be decoupled from the development of data science tools and techniques. More specifically, it will become imperative that data scientists understand how to design algorithms that can operate effectively in environments where data movement is the key performance bottleneck.To meet this need, we will provide training that ensures we generate highly-employable individuals who have both an understanding of the design of future computer hardware as well as an understanding of how and when to flex the algorithmic solutions to best exploit the computational resources that will exist in the future.From the outset, the students will be embedded in a computing environment that anticipates the hardware resources that will arrive on their desks after they graduate, not the hardware that exists today. The cohort of students provides the critical mass that motivates engagement with internationally-leading supercomputing centres: STFC's Hartree Centre is an integral part of the team; links we have established with IBM Research in the US will provide students with access to state-of-the-art computing hardware. This anticipation of future computing capability will ensure our graduates are highly employable, but also help motivate end-user organisations to engage with the CDT.We have identified such end-user organisations that span two themes: defence and security; manufacturing. Organisations in these themes are driven by performance demands and efficiency requirements respectively.We will align the training we provide with the needs of the cohort, the theme and the individual. Each studentship will have two academic supervisors (one aligned with the "Future Computing Systems" and one aligned with moving "Towards a Data-Driven Future") and at least one supervisor from a project partner. This supervisory team will co-define the scope of each studentship. Once the high quality student has been selected and recruited, we will work with the student to define the training that aligns with their needs and the specific demands of the studentship. Our training provision will include the training needs associated with both the "Future Computing Systems" and "Towards a Data-Driven Future" priority areas. We will use guest lectures from, for example, IBM (as used to train Fast Track civil servants) and UC Berkeley to ensure we maximise our graduates' ability to thrive and to become tomorrow's leaders in Distributed Algorithms.