Well-designed interactive data science tools enable both flexible and comprehensive data exploration. While the increasing size and complexity of modern data sets demand computational guidance methods that automatically find potentially important patterns, most scientific investigations are inherently exploratory (meaning many the questions a user may wish to ask of the data are not known beforehand) and require interactive query capabilities. Achieving such a balance, however, is not trivial; it requires the judicious orchestration of human strengths, namely creativity, visual perception, and background knowledge, with the power of computational algorithms and machinery. In this talk, I will describe several interactive data science tools and highlight the human interaction, data visualization, and algorithmic guidance methods in each, as well as the integration of all three components into data-driven, human-directed, machine-enabled frameworks. Because these tools were developed in close collaboration with domain experts, I will also highlight practical applications and lessons learned. I will conclude with an overview of the ORNL Visual Informatics for Science and Technology Advancement (VISTA) Data Exploration Laboratory, which features a unique hands-on laboratory for testing new interactive display media, an interdisciplinary community, and diverse data science expertise that contributes to ORNL’s reputation as a leading institution for advanced data analytics research.
Friday, February 5 at 2:30pm to 3:30pmVirtual Event