Systems and Algorithms for Easy-to-use Federated Data Science

数据科学研究和技术对认知神经科学、个性化营销和政府欺诈等应用中的数据洞察产生了重大影响。在这些应用中，需要在多个科学学科和跨地理边界的各种组织维护的数据集上实现数据驱动的发现。这些应用程序需要创新的方式来交换、共享、处理和可视化大量数据。因此，数据科学家需要访问内部和外部数据源，这些数据源积累了大量地理分布和分散的数据集。定期访问这些数据集，以执行通常以管道方式组织的不同数据集成、提取和分析任务。正如最近在《建设创新者之国》中所报道的那样，加拿大投资5.7亿美元，以提高研究人员获得先进计算和大数据资源的机会。这项资助申请符合这一愿景，即使加拿大保持在技术需求最高的领域之一的最前沿，即数据科学。我的研究项目有一个长期愿景，即开发一个易于使用的联邦数据科学平台，包括支持地理分布式数据集上的数据科学应用的高级算法。该资助申请提出了四个短期目标：(a)数据科学平台的自然语言接口，(b)数据科学版本控制，以增强协作，同时保持准确性和公平性，(c)用于管理和优化地理分布式数据集上的数据科学工作负载的联邦引擎，以及(d)技术转让，将我们的发现应用于神经科学并将我们的系统集成到现有的数据科学平台中。这项拨款申请的重点是知识图和链接网络数据形式的地理分布数据集。我们将利用云资源和并行架构来提供协作工作空间，数据科学家可以在其中协同工作并扩展他们的数据科学工作负载。这项拨款申请旨在推动高度竞争领域的最新技术，弥合数据系统和数据驱动智能之间的差距，从大量地理分布数据集中提取洞察力。拟议的研究及其主要成果将支持加拿大参与全球创新竞赛。数据科学市场的就业正在迅速增长。此申请支持在该领域提供高素质人才的培训，以避免加拿大劳动力市场预期的短缺。该项目计划在数据科学的理论和实践方面培养几名学生，并帮助加拿大的各种创业公司从他们的数据中创造商业价值。整个科学学科的范围包括我们的研究项目有前途的实际应用。这将帮助学生使用真实的用例来演示他们的系统。

Data science research and technology drive a significant impact on creating insights from data in applications, such as cognitive neuroscience, personalized marketing, and government fraud. In these applications, there is a need to enable data-driven discoveries on multiple scientific disciplines, and datasets maintained by various organizations across geographical boundaries. These applications demand innovative ways to exchange, share, process, and visualize huge amounts of data. Consequently, data scientists need access to internal and external data sources that accumulate vast amounts of geo-distributed and decentralized datasets. These datasets are accessed on a regular basis to perform different data integration, extraction, and analytics tasks that are usually organized in a pipeline fashion. As reported recently in "Building a Nation of Innovators", Canada invests $570 million to enhance researchers' access to advanced computing and big data resources. This grant application is aligned with this vision to keep Canada at the forefront of one of the highest-demand sectors in technology, i.e., data science. My research program has a long-term vision to develop an easy-to-use federated data science platform, including advanced algorithms to support data science applications on geo-distributed datasets. This grant application proposes four short-term objectives: (a) natural language interfaces for data science platforms, (b) data science versioning control to enhance collaboration while maintaining accuracy and fairness, (c) a federated engine for managing and optimizing data science workloads over geo-distributed datasets, and (d) technology transfer to apply our findings to neuroscience and integrate our systems into existing data science platforms. This grant application focuses on geo-distributed datasets in the form of knowledge graphs and linked web data. We will leverage cloud resources and parallel architectures to provide a collaborative workspace, in which data scientists can work together and scale up their data science workloads. This grant application aims to advance the state of the art in a highly competitive area that bridges the gap between data systems and data-driven intelligence on extracting insight from a vast pool of geo-distributed datasets. The proposed research and its main outcomes will support Canada in competing in the global innovation race. Employment in the data science market is rapidly growing. This application supports training that provides highly qualified personnel in this area to avoid anticipated shortage in the Canadian labour market. This application plans to train several students in the theoretical and practical aspects of data science, and help various Canadian startups to create business value from their data. The whole spectrum of scientific disciplines includes promising real applications for our research program. This will help the students to demonstrate their systems using real use-cases.