SenVis: Interactive Tensor?based Sensitivity Visualization

Sobol's method is one of the most powerful and widely used frameworks for global sensitivity analysis, it maps every possible combination input variables to an associated Sobol index. However, these indices are often challenging analyze in depth, due part lack suitable, flexible enough, fast-to-query data access structures as well visualization techniques. We propose a tool that leverages tensor decomposition, compressed format can quickly approximately answer sophisticated queries over exponential-sized sets indices. This way, we able capture complete information high-dimensional scalar models. Our application based on three-stage visualization, which be analyzed added or removed interactively. It includes novel hourglass-like diagram presenting relative importance any single variable with respect composition rest variables. showcase our range example models, whereby demonstrate high expressive power analytical capability made proposed method.