An approach to the perceptual optimization of complex visualizations

摘要: This paper proposes a new experimental framework within which evidence regarding the perceptual characteristics of visualization method can be collected, and describes how this explored to discover principles insights guide design perceptually near-optimal visualizations. We make case that each current approaches for evaluating visualizations is limited in what it tell us about optimal tuning visual design. go on argue our approach better suited optimizing kinds complex displays are commonly created visualization. Our uses human-in-the-loop experiments selectively search through parameter space method, generating large databases rated solutions. Data mining then used extract results from database, ranging highly specific exemplar particular data set, more broadly applicable guidelines illustrate using recent study texturing layered surfaces viewed stereo motion. show genetic algorithm valuable way guiding space. also demonstrate several useful methods including clustering, principal component analysis, neural networks, statistical comparisons functions parameters.