Computing and networking, data collection and analytics have been booming in the recent decades to manifest the coming Big Data era. Nowadays, more and more data are being collected and analyzed by organizations to make critical decisions. However, many challenges remain to be addressed.
One challenge is that it is not easy to track which versions of a data set flow through which versions of the cleaning, transformation and analytics to produce the decisions, especially with constantly updating heterogeneous data sources, complex human-in-the loop data wrangling, and sometimes non-reproducible black-box data analytics and version control of big data;
Another challenge is to support data scientists to easily explore subsets of data locally, to share reproducible versions of their exploration in the team, and finally to transform the successful versions into large-scale deployment for continuously serving the learned insights and models of other decision-making systems.
Yet another challenge is to keep semantically and physically consistence of views among distributed and multiple data stores. This can be seen from many real-life examples. In the machine learning and data mining community, researchers require data to data and data to model links with proper provenance of information. In the scientific computing community, smart and efficient management of large amounts of data going through various computation workflows with some degree of reproducibility is also a must. In large scale distributed systems such as clouds, data management tends to employ multi-store systems where global data are split based on various criteria and then stored into distributed multiple stores, thus making the maintenance of consistency difficult.
In this workshop, we bring together researchers and practitioners to share their novel approaches and experiences of managing data to decision “ pipelines” in a production or exploration environment, the transparency and trustworthiness of decision and data, data processing and tracking for closing the gap between advanced computing and platforms, and data management and engineering for newly emerging challenges and topics.