Cieran: Designing Sequential Colormaps via In-Situ Active Preference Learning

Abstract

Figure 1: Cieran's preference learning interface used with Altair [71]. Cieran supports efficient colormap selection within an analyst's workflow through integration with Jupyter Notebooks. Cieran first interpolates example colormaps through a chosen color (e.g., #186E8D, a teal blue shown in the progress bar). People iteratively input preferences to Cieran by making value judgements across pairs of examples, and Cieran uses the preference data to induce a context-specific model of aesthetic utility. This model used to rank and create colormaps.