Workshop: Handling uncertainties in big data (HUBD)

The application of machine learning in big data has attracted a lot of attention, both in science and in business. Society and business increasingly depend on data analytics as a basis for decision-making. While almost every practitioner is familiar with the ‘garbage in = garbage out’ adage, there are many sources of uncertainties that are not always taken into account, not widely understood, and especially are not studied in relation to each other. Sources can be inaccurate, noisy or (temporarily) unavailable data; poor or faulty preprocessing and filtering; models that are trained on non-representative or biased datasets; or inadequate visualizations and interpretations of the output.

The goal of this workshop is to create insights in the relations between all sources of uncertainty in big data processing pipelines, to create a more comprehensive approach that enables business to deploy reliable, accurate processing. The workshop aims to create an overview of the state of the art in handling uncertainties in big data, and to identify the areas that need more attention in order to create reliable, useful and acceptable Big Data applications.

Call for papers

Contributions are expected from research that provides original and new insights into the effects of combined uncertainties in big data processing pipelines, including, but not limited to, research topics like:

• How do uncertainties propagate through complex big data processing pipelines?
• How can mismatches between training data and operational data be detected and mitigated in operation?
• How can data quality of data streams in multistakeholder collaboration data processing environments be measured and maintained?
• How to communicate uncertainties to end users in such a way that they have an optimal understanding of the reliability of the output?
• How to design an overall data processing pipeline in such a way that the uncertainties are optimized in relation to the user objective function?
• How can the level of uncertainty be measured for machine learning approaches that do not allow inspection (like deep learning)?
• How can semantic uncertainties be combined with quantitative uncertainties?

The workshop will take place during the afternoon session on the first day of the 14th IEEE eScience 2018 event.

Are you a scientist, or are you interested in improving your business or product? Make sure to join our workshop!