We propose the concept of Speculative Execution for Visual Analytics and discuss its effectiveness for model exploration and optimization. Speculative Execution enables the automatic generation of alternative, competing model configurations that do not alter the current model state unless explicitly confirmed by the user. These alternatives are computed based on either user interactions or model quality measures and can be explored using delta-visualizations. By automatically proposing modeling alternatives, systems employing Speculative Execution can shorten the gap between users and models, reduce the confirmation bias and speed up optimization processes. In this paper, we have assembled five application scenarios showcasing the potential of Speculative Execution, as well as a potential for further research.
Speculative Execution (SpecEx) can be retrofitted to existing VA systems that can either collect model quality metrics or record user interaction data. These data can then be used as a foundation to trigger SpecEx sandboxes that explore different modeling alternatives. Effective, bespoke Delta Visualizations are needed for each model to highligh the differences between sandboxes and the “original” model. Users can accept or reject any of the sandboxes and continue the visual analytics process from there.
We have implemented a Visual Analytics framework for the user-steerable optimization of topic models on streaming text.
See the video accompanying our VAST’18 paper below: